Update repo #9

Update INSTALL.md

Add ISSUE_TEMPLATE for Github

Fix a small typo

It is simpler and requires less effort (i.e. no auth) compared to SSH

Access repos via https in INSTALL.md

* Add missing __init__.py files * Add packages * Rename logging.py to logger.py Import rules from Python2 makes this a bad idea * Make import_file py2 compatible * list does not have .copy() in py2 * math.log2 does not exist in py2 * Miscellaneous fixes for py2 * Address comments

Fix BoxList constructor call in README.md

Replace baseline with benchmark

Replace BoxList argument size with image_size.

use cuda version torch.randperm to avoid copy from gpu to cpu and a fatal bug in multi-thread cpu version

* quick fix for area * Update bounding_box.py

## Bug When I repeatedly invoke build_detection_model function to build the model, I will get the error: `AttributeError: 'ResNet' object has no attribute 'layer1'`. ## To Reproduce ``` model = build_detection_model(cfg) model = build_detection_model(cfg) ``` ## The reason The variable `ResNet50StagesTo4` is a global generator expression, so when we build the model secondly, the code executes to Line 82 `for stage_spec in stage_specs:` and `stage_specs` will return empty leading to do not add any stage. Finally, `self._freeze_backbone` try to freeze the backbone by executing `m = getattr(self, "layer" + str(stage_index))` firstly. At the moment, it will throw the AttributeError `AttributeError: 'ResNet' object has no attribute 'layer1'`. I guess you want to define ResNet50StagesTo4 as the tuple, so I try to fix by add tuple type qualifier. ## The solution Add the tuple type to `ResNet50StagesTo5`, `ResNet50StagesTo4`, `ResNet50FPNStagesTo5`, `ResNet101FPNStagesTo5`. I do not know whether there are similar bug existing, so you need to review my solution. Thank you!

* added Dockerfile * intructions for installation with docker * adds nccl to dockerfile * Update Dockerfile * Update INSTALL.md

* adding registry to hook custom building blocks * adding customizable rpn head * support customizable c2 weight loading

* added Dockerfile * intructions for installation with docker * INSTALL.md: added note to inform nvidia-docker only works with Linux * dockerfile nccl2 hotfix * INSTALL.md update * removes explicit install of NCCL2 as it is included in the CUDA image

* fix some minor issues in dockerfile RUN echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections fixes debconf: unable to initialize frontend: Dialog debconf: (TERM is not set, so the dialog frontend is not usable.) debconf: falling back to frontend: Readline debconf: unable to initialize frontend: Readline debconf: (This frontend requires a controlling tty.) debconf: falling back to frontend: Teletype dpkg-preconfigure: unable to re-open stdin: RUN apt-get update -y && apt-get install -y apt-utils fixes debconf: delaying package configuration, since apt-utils is not installed * remove redundant command * add -y to conda install and pip install * remove -y from pip install

* Add an option to postprocess masks during inference * Fix COCO evaluation to resize masks ony if needed. * Fix casting * Fix minor issues in paste_mask_in_image * Cast mask to uint8 * Make Masker batch compatible * Remove warnings and stylistic changes

* add opencv support to Dockerfile cv2 is used by predictor.py, so it's required * install opencv-python from conda

* add support for pascal voc dataset and evaluate * optimization for adding voc dataset * make inference.py dataset-agnostic; add use_difficult option to voc dataset * handle voc difficult objects correctly * Remove dependency on lxml plus minor improvements * More cleanups * More comments and improvements * Lint fix * Move configs to their own folder

* Dockerfile with jupyter notebook support * add instruction of building docker with jupyter * upload dockerfile with jupyter support * remove Dockerfile-jupyter * remove jupyter_notebook_config.py * update INSTALL.md * Dockerfile with jupyter notebook support * add instruction of building docker with jupyter * upload dockerfile with jupyter support * remove Dockerfile-jupyter * remove jupyter_notebook_config.py * update INSTALL.md * apply changes in 1123 * update INSTALL.md * update install.md add description of the password add -v flag * clean up * remove tensorflow copyright

* fix the same issue as #185 * add fix from the previous change

In maskrcnn_benchmark/modeling/roi_heads/mask_head/inference.py change line 125-126 to w = int(box[2] - box[0] + TO_REMOVE) h = int(box[3] - box[1] + TO_REMOVE)

* readme and predictor.py * format modification and readme

…stributed (#248) * make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed

* add force json option * fix the same issue as #185 * bug fix * cityscapes config * update paths catalog * discard config change * organize code for more-datasets * use better representation for coco-style datasets * rename coco-style config * remove import * chmod 644 * make the config more verbose * update readme * rename * chmod

* Fix indexing error in nms kernel Here it is indexing a cuda tensor with CPU indices. It used to work, but after pytorch/pytorch@006505b it results in memory corruption. * Use the device of other tensors

* Update checkpoint.py if edit "{save_dir}/last_checkpoint" by vim, "\n" will auto append to the end of file. which will raise ``` FileNotFoundError: [Errno 2] No such file or directory: 'save_dir/model_0060000.pth\n' ``` * Update checkpoint.py if edit "{save_dir}/last_checkpoint" by vim, "\n" will auto append to the end of file. which will raise ``` FileNotFoundError: [Errno 2] No such file or directory: 'save_dir/model_0060000.pth\n' ```

sort keys before reduction

* make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication

This reverts commit 6cbb3d2.

…atibility (#354) fixes #339

…371) * Add new section "Projects using maskrcnn-benchmark". * Update README.md update the format. * Update README.md

* Add new section "Projects using maskrcnn-benchmark". * Update README.md update the format. * Update README.md * Add coco_2017_train and coco_2017_val * Update README.md Add the instructions about COCO_2017

…#377) * make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication * use 'kaiming_uniform' to initialize resnet, disable gn after fc layer, and add dilation into ResNetHead

Since IMS_PER_BATCH is global, it shouldn't be multiplied with the # of GPUs.

… R-CNN (#405)

* make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication * use 'kaiming_uniform' to initialize resnet, disable gn after fc layer, and add dilation into ResNetHead * agnostic-regression for bbox

…407)

…eilDiv. Adding type-cast to (long) to first parameter to fix the errors. (#409)

* Registry for RoI Box Predictors - Add a registry ROI_BOX_PREDICTOR - Use the registry in roi_box_predictors.py, replacing the local factory - Minor changes in structures/bounding_box.py: when copying a box with fields, check if the field exists - Minor changes in logger.py: make filename a optional argument with default value of "log.txt" * Add Argument skip_missing=False

@ChangErgou

* [WIP] Keypoints inference on C2 models work * Training seems to work Still gives slightly worse results * e2e training works but gives 3 and 5 mAP less * Add modification proposed by @ChangErgou Improves mAP by 1.5 points, to 0.514 and 0.609 * Keypoints reproduce expected results * Clean coco.py * Linter + remove unnecessary code * Merge criteria for empty bboxes in has_valid_annotation * Remove trailing print * Add demo support for keypoints Still need further cleanups and improvements, like adding fields support for the other ops in Keypoints * More cleanups and misc improvements * Fixes after rebase * Add information to the readme * Fix md formatting

* Add RPN config files * Add more RPN models

* Adding support to Caffe2 ResNeXt-152-32x8d-FPN-IN5k backbone for Mask R-CNN * Clean up * Fixing path_catalogs.py * Back to old ROIAlign_cpu.cpp file

* Add RetinetNet parameters in cfg. * hot fix. * Add the retinanet head module now. * Add the function to generate the anchors for RetinaNet. * Add the SigmoidFocalLoss cuda operator. * Fix the bug in the extra layers. * Change the normalizer for SigmoidFocalLoss * Support multiscale in training. * Add retinannet training script. * Add the inference part of RetinaNet. * Fix the bug when building the extra layers in retinanet. Update the matching part in retinanet_loss. * Add the first version of the inference of RetinaNet. Need to check it again to see if is there any room for speed improvement. * Remove the retinanet_R-50-FPN_2x.yaml first. * Optimize the retinanet postprocessing. * quick fix. * Add script for training RetinaNet with ResNet101 backbone. * Move cfg.RETINANET to cfg.MODEL.RETINANET * Remove the variables which are not used. * revert boxlist_ops. Generate Empty BoxLists instead of [] in retinanet_infer * Remove the not used commented lines. Add NUM_DETECTIONS_PER_IMAGE * remove the not used codes. * Move retinanet related files under Modeling/rpn/retinanet * Add retinanet_X_101_32x8d_FPN_1x.yaml script. This model is not fully validated. I only trained it around 5000 iterations and everything is fine. * set RETINANET.PRE_NMS_TOP_N as 0 in level5 (p7), because previous setting may generate zero detections and could cause the program break. This part is used in original Detectron setting. * Fix the rpn only bug when the training ends. * Minor improvements * Comments and add Python-only implementation * Bugfix and remove commented code * keep the generalized_rcnn same. Move the build_retinanet inside build_rpn. * Add USE_C5 in the MODEL.RETINANET * Add two configs using P5 to generate P6. * fix the bug when loading the Caffe2 ImageNet pretrained model. * Reduce the code depulication of RPN loss and RetinaNet loss. * Remove the comment which is not used. * Remove the hard coded number of classes. * share the foward part of rpn inference. * fix the bug in rpn inference. * Remove the conditional part in the inference. * Bug fix: add the utils file for permute and flatten of the box prediction layers. * Update the comment. * quick fix. Adding import cat. * quick fix: forget including import. * Adjust the normalization part according to Detectron's setting. * Use the bbox reg normalization term. * Clean the code according to recent review. * Using CUDA version for training now. And the python version for training on cpu. * rename the directory to retinanet. * Make the train and val datasets are consistent with mask r-cnn setting. * add comment.

A few addition: I added the top level directory `cityscapes` since the `tools/cityscapes/convert_cityscapes_to_coco.py` script has the directory structure `gtFine_trainvaltest/gtFine` hardcoded into it which is fine but was not clear at first. Also added a **Note** to warn people to install detectron as well, since the script uses `detectron.utils.boxes` and `detectron.utils.segm` modules which has further dependencies in the detectron lib.

The previous instruction examples assumed that the directory `~/github` existed, and did not include any check to create it if the directory did not exist. I updated to install in whatever directory the user is current in. I also updated to make it clear how the CUDA version is specified, and fixed a typo in activating the conda env.

the env variable is misused in the current version

Finishing the clean up in #455, unsetting the proper variable. In general, thanks for making this so easy to install! I'd run into all kinds of versioning issues (version of Ubuntu not playing nicely with versions of CUDA/pytorch/libraries) trying to install other libraries implementing these algorithms. I'm super impressed by the quality of support, and the easy install, for this library.

* Remove Detectron dependency I have looked into the boxes.py to swap [these lines](https://github.com/facebookresearch/Detectron/blob/8170b25b425967f8f1c7d715bea3c5b8d9536cd8/detectron/utils/boxes.py#L51L52): ``` import detectron.utils.cython_bbox as cython_bbox import detectron.utils.cython_nms as cython_nms ``` ``` from maskrcnn_benchmark.structures.boxlist_ops import boxlist_iou from maskrcnn_benchmark.structures.boxlist_ops import boxlist_nms ``` However some functions are missing from the `boxlist_ops` like the [`soft_nms`](https://github.com/facebookresearch/Detectron/blob/master/detectron/utils/cython_nms.pyx#L98L203) . So I just tried to modify the `maskrcnn-benchmark/tools/cityscapes/convert_cityscapes_to_coco.py` script. Here we have `polys_to_boxes` function from `segms.py` and I could not find its analogous in the maskrcnn_benchmark lib. It seems to me that the original function in `segms.py` is using pure lists so I just wrote two auxiliary functions reusing the boxList's convert method( https://github.com/facebookresearch/maskrcnn-benchmark/blob/master/maskrcnn_benchmark/structures/bounding_box.py#L67L70 ) and Detectron's polys_to_boxes ( https://github.com/facebookresearch/Detectron/blob/b5dcc0fe1d091cb70f9243939258215dd63e3dfa/detectron/utils/segms.py#L135L140 ): ``` def poly_to_box(poly): """Convert a polygon into a tight bounding box.""" x0 = min(min(p[::2]) for p in poly) x1 = max(max(p[::2]) for p in poly) y0 = min(min(p[1::2]) for p in poly) y1 = max(max(p[1::2]) for p in poly) box_from_poly = [x0, y0, x1, y1] return box_from_poly def xyxy_to_xywh(xyxy_box): xmin, ymin, xmax, ymax = xyxy_box TO_REMOVE = 1 xywh_box = (xmin, ymin, xmax - xmin + TO_REMOVE, ymax - ymin + TO_REMOVE) return xywh_box ``` * removed leftovers * Update convert_cityscapes_to_coco.py

I'm using pycharm to debug the code on a remote server, the remote debugging seems to be performed by pytest and it pops errors like: train_net.py E test setup failed file /tmp/pycharm_project_269/tools/train_net.py, line 79 def test(cfg, model, distributed): E fixture 'cfg' not found > available fixtures: cache, capfd, capfdbinary, caplog, capsys, capsysbinary, doctest_namespace, monkeypatch, pytestconfig, record_property, record_xml_attribute, recwarn, tmp_path, tmp_path_factory, tmpdir, tmpdir_factory > use 'pytest --fixtures [testpath]' for help on them. it seems the function name ‘test()’ has come conflict with pytest, so it may be better use another name.

* make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication * use 'kaiming_uniform' to initialize resnet, disable gn after fc layer, and add dilation into ResNetHead * agnostic-regression for bbox * please set 'STRIDE_IN_1X1' to be 'False' when backbone use GN

* Supported any feature map size for average pool. * Different models may have different feature map size. * Used registry to register keypoint and mask heads. * Passing in/out channels between modules when creating the model. Passing in/out channels between modules when creating the model. This simplifies the code to compute the input channels for feature extractors and makes the predictors independent of the backbone architectures. * Passed in_channels to rpn and head builders. * Set out_channels to model modules including backbone and feature extractors. * Moved cfg.MODEL.BACKBONE.OUT_CHANNELS to cfg.MODEL.RESNETS.BACKBONE_OUT_CHANNELS as it is not used by all architectures. Updated config files accordingly. For new architecture modules, the return module needs to contain a field called 'out_channels' to indicate the output channel size. * Added unit test for box_coder and nms. * Added FBNet architecture. * FBNet is a general architecture definition to support efficient architecture search and MaskRCNN2GO. * Included various efficient building blocks (inverted residual, shuffle, separate dw conv, dw upsampling etc.) * Supported building backbone, rpn, detection, keypoint and mask heads using efficient building blocks. * Architecture could be defined in `fbnet_modeldef.py` or in `cfg.MODEL.FBNET.ARCH_DEF` directly. * A few baseline architectures are included. * Added various unit tests. * build and run backbones. * build and run feature extractors. * build and run predictors. * Added a unit test to verify all config files are loadable.

There were two `RESNETS` sections, which overrode each other, leading to error messages like: ``` RuntimeError: Error(s) in loading state_dict for GeneralizedRCNN: size mismatch for backbone.fpn.fpn_inner1.weight: copying a param with shape torch.Size([256, 256, 1, 1]) from checkpoint, the shape in current model is torch.Size([1024, 256, 1, 1]). ... size mismatch for roi_heads.mask.feature_extractor.mask_fcn1.weight: copying a param with shape torch.Size([256, 256, 3, 3]) from checkpoint, the shape in current model is torch.Size([256, 1024, 3, 3]). ``` This just combines them back into a single section, while maintaining all param values. That got the model running again for me.

* make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication * use 'kaiming_uniform' to initialize resnet, disable gn after fc layer, and add dilation into ResNetHead * agnostic-regression for bbox * please set 'STRIDE_IN_1X1' to be 'False' when backbone use GN * add README.md for GN

* Add new section "Projects using maskrcnn-benchmark". * Update README.md update the format. * Update README.md * Add coco_2017_train and coco_2017_val * Update README.md Add the instructions about COCO_2017 * Update the pip install. Adding tqdm which is used in engine/inference.py

…buted' has no attribute 'is_initialized'. (#511)

* fix INSTALL.md * fix PR * Update INSTALL.md

* Added a timer to benchmark model inference time in addition to total runtime. * Updated FBNet configs and included some baselines benchmark results. * Added a unit test for detectors. * Add links to the models

Fix dispatch breakage

* support for any one cuda device * Revert "support for any one cuda device" This reverts commit 0197e4e. * support runnning for anyone cuda device * using safe CUDAGuard rather than intrinsic CUDASetDevice * supplement a header dependency (test passed) * Support for arbitrary GPU device. * Support for arbitrary GPU device. * add docs for two method to control devices

* Merge branch 'master' of /home/braincreator/projects/maskrcnn-benchmark with conflicts. * rolls back the breaking AT dispatch changes (#555) * revert accidental docker changes * revert accidental docker changes (2)

* fixes to dockerfile * replaces local installation by git clone

This reverts commit f031879.

@miguelvr

…cker (#612) * add a FORCE_CUDA flag Following discussion [here](#167), this seemed the best solution * Update Dockerfile * Update setup.py * add FORCE_CUDA as an ARG * modified: docker/Dockerfile modified: setup.py * small fix to readme of demo * remove test print * keep ARG_CUDA * remove env value and use the one from ARG * keep same formatting as source * change proposed by @miguelvr * Update INSTALL.md

Fix a bug. Romove the echo command in line 36: RUN conda install pytorch-nightly cudatoolkit=${CUDA} -c pytorch To enable conda installation of pytorch-nightly.

add tqdm in line32 : RUN pip install ninja yacs cython matplotlib opencv-python tqdm

* support RLE and binary mask * do not convert to numpy * be consistent with Detectron * delete wrong comment * [WIP] add tests for segmentation_mask * update tests * minor change * Refactored segmentation_mask.py * Add unit test for segmentation_mask.py * Add RLE support for BinaryMaskList * PEP8 black formatting * Minor patch * Use internal that handles 0 channels * Fix polygon slicing

* fix py2 * fix py2

https://gist.github.com/botcs/a59f3f59e22e5df93e3e5e4f86718af3

* make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication * use 'kaiming_uniform' to initialize resnet, disable gn after fc layer, and add dilation into ResNetHead * agnostic-regression for bbox * please set 'STRIDE_IN_1X1' to be 'False' when backbone use GN * add README.md for GN * add dcn from mmdetection

* add color jitter augmentation * fix spelling

* proposals from RPN per image during training * README * Update README for setting FPN_POST_NMS_TOP_N_TRAIN * Update README.md * removing extra space change

@fmassa

* Initial multi-precision training Adds fp16 support via apex.amp Also switches communication to apex.DistributedDataParallel * Add Apex install to dockerfile * Fixes from @fmassa review Added support to tools/test_net.py SOLVER.MIXED_PRECISION -> DTYPE \in {float32, float16} apex.amp not installed now raises ImportError * Remove extraneous apex DDP import * Move to new amp API

* make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication * use 'kaiming_uniform' to initialize resnet, disable gn after fc layer, and add dilation into ResNetHead * agnostic-regression for bbox * please set 'STRIDE_IN_1X1' to be 'False' when backbone use GN * add README.md for GN * add dcn from mmdetection * add documentation for finetuning cityscapes * add documentation for finetuning cityscapes * add documentation for finetuning cityscapes

…ME.md' (#701) * make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication * use 'kaiming_uniform' to initialize resnet, disable gn after fc layer, and add dilation into ResNetHead * agnostic-regression for bbox * please set 'STRIDE_IN_1X1' to be 'False' when backbone use GN * add README.md for GN * add dcn from mmdetection * add documentation for finetuning cityscapes * add documentation for finetuning cityscapes * add documentation for finetuning cityscapes * add 'once_differentiable' for dcn and modify 'configs/cityscapes/README.md'

* Add ImportError to generic except * Black formatter * OpenCV is no longer optional * Update README w.r.t. FPN_POST_NMS_PER_BATCH

* Update documentation for testing and calculating mAP for each class * Update documentation for testing and calculating mAP for each class

* make pixel indexes 0-based for bounding box in pascal voc dataset * replacing all instances of torch.distributed.deprecated with torch.distributed * replacing all instances of torch.distributed.deprecated with torch.distributed * add GroupNorm * add GroupNorm -- sort out yaml files * use torch.nn.GroupNorm instead, replace 'use_gn' with 'conv_block' and use 'BaseStem'&'Bottleneck' to simply codes * modification on 'group_norm' and 'conv_with_kaiming_uniform' function * modification on yaml files in configs/gn_baselines/ and reduce the amount of indentation and code duplication * use 'kaiming_uniform' to initialize resnet, disable gn after fc layer, and add dilation into ResNetHead * agnostic-regression for bbox * please set 'STRIDE_IN_1X1' to be 'False' when backbone use GN * add README.md for GN * add dcn from mmdetection * add documentation for finetuning cityscapes * add documentation for finetuning cityscapes * add documentation for finetuning cityscapes * add 'once_differentiable' for dcn and modify 'configs/cityscapes/README.md' * rectify the 'padding' for 'DFConv2d'

According to the docs: https://pytorch.org/docs/stable/optim.html#torch.optim.lr_scheduler.StepLR, now we need to call `scheduler.step()` after `optimizer.step()`. Breaking change PR: pytorch/pytorch#7889

Avoids naming clash with torchvision CocoDetection class

…umel (#799)

* modify senario where dataset has no mask provided * change to check segmentation mask in the way of keypoint

* max proposal protect and box convert to xyxy in iou function * remove unnecessary lines

* Update SigmoidFocalLoss_cuda.cu line 120 converted to long to support windows10 * Update SigmoidFocalLoss_cuda.cu removed old line and converted to long to support windows10 * Update SigmoidFocalLoss_cuda.cu Converted to long to support windows10 * Update INSTALL.md Step_By_Step Guide installation on windows * Update INSTALL.md updated Clone of architecture to be the one of this repo which has the fixes needed for Windows10 * Update INSTALL.md How to check what cuda version to install with pytorch * Update INSTALL.md

* Implement multi-scale testing(bbox aug) like Detectron. * Add comment. * Fix missing cfg after merge.

* fix the bug in segmentation_mask when initializing BinaryMaskList with an empty list * Add empty binary mask handling * Strict indexing in BinaryMaskList - raise error for empty indexing

remove `_C.MODEL.BACKBONE.USE_GN = False`, since it is not used in the code. ``` liliang@liliang-ubuntu:~/liliang_learning/maskrcnn-benchmark/maskrcnn_benchmark$ grep -r "USE_GN" Binary file config/__pycache__/defaults.cpython-37.pyc matches config/defaults.py:_C.MODEL.BACKBONE.USE_GN = False config/defaults.py:_C.MODEL.FPN.USE_GN = False config/defaults.py:_C.MODEL.ROI_BOX_HEAD.USE_GN = False config/defaults.py:_C.MODEL.ROI_MASK_HEAD.USE_GN = False Binary file modeling/backbone/__pycache__/backbone.cpython-37.pyc matches modeling/backbone/backbone.py: cfg.MODEL.FPN.USE_GN, cfg.MODEL.FPN.USE_RELU modeling/backbone/backbone.py: cfg.MODEL.FPN.USE_GN, cfg.MODEL.FPN.USE_RELU Binary file modeling/roi_heads/mask_head/__pycache__/roi_mask_feature_extractors.cpython-37.pyc matches modeling/roi_heads/mask_head/roi_mask_feature_extractors.py: use_gn = cfg.MODEL.ROI_MASK_HEAD.USE_GN Binary file modeling/roi_heads/box_head/__pycache__/roi_box_feature_extractors.cpython-37.pyc matches modeling/roi_heads/box_head/roi_box_feature_extractors.py: use_gn = cfg.MODEL.ROI_BOX_HEAD.USE_GN modeling/roi_heads/box_head/roi_box_feature_extractors.py: use_gn = cfg.MODEL.ROI_BOX_HEAD.USE_GN ```

… time (#798)

* Add requests for compatibility with the demo notebook * Add requests for compatibility with the demo notebook

* do not split strings so that format() works as expected * address flake8 indentation issue

Fixed missing line importing F (torchvision.transforms.functional) used in line 43

now can inference on cpu by setting MODEL.DEVICE cpu

* keep the resize function the same in test time the same with training time * add vertical flip

@miguelvr

…f the images during training` (#836) * Update setup.py * add a FORCE_CUDA flag Following discussion [here](#167), this seemed the best solution * Update Dockerfile * Update setup.py * add FORCE_CUDA as an ARG * modified: docker/Dockerfile modified: setup.py * small fix to readme of demo * remove test print * keep ARG_CUDA * remove env value and use the one from ARG * keep same formatting as source * change proposed by @miguelvr * Update INSTALL.md * Update trainer.py * Update trainer.py * Update trainer.py * Update trainer.py * Update trainer.py * Update setup.py

* Merge branch 'master' of /home/braincreator/projects/maskrcnn-benchmark with conflicts. * update Dockerfile * save config in output dir * replace string format with os.path.join

…flake8 fix (#846)

* Merge branch 'master' of /home/braincreator/projects/maskrcnn-benchmark with conflicts. * update Dockerfile * save labels to output dir * save labels on main process only

* Add mixed precision training instruction Highlight the model support for mixed precision training and add instruction to enable it on README * Add link to in-depth MP training guide

* support RLE and binary mask * do not convert to numpy * be consistent with Detectron * delete wrong comment * [WIP] add tests for segmentation_mask * update tests * minor change * Refactored segmentation_mask.py * Add unit test for segmentation_mask.py * Add RLE support for BinaryMaskList * PEP8 black formatting * Minor patch * Use internal that handles 0 channels * Fix polygon slicing * Revert "Pr150" * Add RLE size check for SegmentationMask

Adding weight loading from saved pth file.

number of instances should be in the first dimension.

Fix typo, highlight words in python language, add hyperlink

From https://github.com/facebookresearch/maskrcnn-benchmark/blob/master/INSTALL.md: 1. additionally install tqdm, cudatoolkit, apex 2. add FORCE_CUDA argument

* print pretty evaluation results * remove `name` attribute from Datasets * add a newline before printing results

* Added panoptic segmentation demo * Removed overlong notebook outputs

…pes it as needed

creating the conda env does not come with a local version of pip, therefore in line 27 it would be quite non-deterministic where the installed lib would end up, possibly causing collisions with different versions of this or other repos

Update segmentation_mask.py

Add Abstract dataset definition

…gamma_alpha gamma, alpha are supposed to be floats, delete lines with [0] subscripts #1013

Need this)

fix minor typos

install local pip with conda

Update INSTALL.md

Add uncompressed rle condition in Binarymask

Update predictor.py

remove redundant reshape of box_regression

Update lr scheduling to pytorch 1.1.0

Validation during training (version 2)

…r is deprecated in pytorch v1.2.0

replacing dtype torch.uint8 with torch.bool for indexing in pytorch 1.2.0

Add native CityScapes evaluation tool

Extend COCO evaluation for AbstractDataset

bugfix: use correct config for tta and device handling during inference

Add native CityScapes dataset

Move horizontal flip probability to config

add the MULAN project in readme.MD, "Projects using maskrcnn-benchmark"

@fmassa

Thanks for your excellent work! @fmassa Based on this repository, we research on the object detectors without sampling heuristics (e.g., Focal Loss, GHM, Undersampling). The paper (https://arxiv.org/abs/1909.04868) and code (https://github.com/ChenJoya/sampling-free) have been released. Thank you again for maskrcnn-benchmark. It is a really simple, efficient, high performance object detection benchmark.

Commits on Oct 29, 2018

Fix weight init in FastRCNNPredictor (#60 )

fmassa authored Oct 29, 2018

Configuration menu

View commit details

Copy full SHA for 7282f55

Browse repository at this point

Copy the full SHA

7282f55 View commit details

Browse the repository at this point in the history

Commits on Nov 20, 2018

Fix demo after PostProcess mask (#187 )

fmassa authored Nov 20, 2018

Configuration menu

View commit details

Copy full SHA for 6855f23

Browse repository at this point

Copy the full SHA

6855f23 View commit details

Browse the repository at this point in the history

Commits on Mar 1, 2019

typo fixed (#517 )

Erik authored and fmassa committed Mar 1, 2019

Configuration menu

View commit details

Copy full SHA for 4a21d69

Browse repository at this point

Copy the full SHA

4a21d69 View commit details

Browse the repository at this point in the history

Commits on Apr 10, 2019

fix #258 (#648 )

lufficc authored and fmassa committed Apr 10, 2019

Configuration menu

View commit details

Copy full SHA for 59d445d

Browse repository at this point

Copy the full SHA

59d445d View commit details

Browse the repository at this point in the history

Commits on Jul 1, 2019

Update loss.py

SkeletonOne authored Jul 1, 2019

Configuration menu

View commit details

Copy full SHA for 9cbae8e

Browse repository at this point

Copy the full SHA

9cbae8e View commit details

Browse the repository at this point in the history

Commits on Sep 20, 2019

NUM_CLASSES: 9->11; Small import fix

botcs committed Sep 20, 2019

Configuration menu

View commit details

Copy full SHA for b9547f4

Browse repository at this point

Copy the full SHA

b9547f4 View commit details

Browse the repository at this point in the history

Update repo #9

Are you sure you want to change the base?

Update repo #9

Commits on Oct 25, 2018

Commits on Oct 26, 2018

Commits on Oct 29, 2018

Commits on Oct 31, 2018

Commits on Nov 1, 2018

Commits on Nov 2, 2018

Commits on Nov 4, 2018

Commits on Nov 5, 2018

Commits on Nov 13, 2018

Commits on Nov 14, 2018

Commits on Nov 16, 2018

Commits on Nov 17, 2018

Commits on Nov 19, 2018

Commits on Nov 20, 2018

Commits on Nov 21, 2018

Commits on Nov 22, 2018

Commits on Nov 23, 2018

Commits on Nov 26, 2018

Commits on Nov 27, 2018

Commits on Nov 28, 2018

Commits on Dec 4, 2018

Commits on Dec 5, 2018

Commits on Dec 6, 2018

Commits on Dec 14, 2018

Commits on Dec 17, 2018

Commits on Dec 18, 2018

Commits on Dec 20, 2018

Commits on Dec 21, 2018

Commits on Dec 30, 2018

Commits on Jan 14, 2019

Commits on Jan 21, 2019

Commits on Jan 22, 2019

Commits on Jan 23, 2019

Commits on Jan 25, 2019

Commits on Jan 30, 2019

Commits on Jan 31, 2019

Commits on Feb 4, 2019

Commits on Feb 5, 2019

Commits on Feb 6, 2019

Commits on Feb 12, 2019

Commits on Feb 13, 2019

Commits on Feb 14, 2019

Commits on Feb 15, 2019

Commits on Feb 18, 2019

Commits on Feb 19, 2019

Commits on Feb 20, 2019

Commits on Feb 22, 2019

Commits on Feb 26, 2019

Commits on Feb 28, 2019

Commits on Mar 1, 2019

Commits on Mar 5, 2019

Commits on Mar 7, 2019

Commits on Mar 9, 2019

Commits on Mar 11, 2019

Commits on Mar 12, 2019

Commits on Mar 25, 2019

Commits on Mar 26, 2019

Commits on Mar 31, 2019

Commits on Apr 2, 2019

Commits on Apr 4, 2019

Commits on Apr 9, 2019

Commits on Apr 10, 2019

Commits on Apr 11, 2019

Commits on Apr 13, 2019

Commits on Apr 16, 2019

Commits on Apr 19, 2019

Commits on Apr 20, 2019

Commits on Apr 22, 2019

Commits on Apr 23, 2019

Commits on Apr 25, 2019

Commits on May 1, 2019

Commits on May 2, 2019

Commits on May 20, 2019

Commits on May 24, 2019

Commits on May 25, 2019

Commits on May 27, 2019

Commits on May 28, 2019