Table with samples (#11)

* done

* minor changes

README.md

@@ -2,6 +2,16 @@
 
 ![rf100blog-mosaicthing](https://user-images.githubusercontent.com/15908060/202452898-9ca6b8f7-4805-4e8e-949a-6e080d7b94d2.jpg)
 
+This repository implements the Roboflow 100 benchmark developed by [Roboflow](https://roboflow.com/). It contains code to download the dataset and reproduce
+mAP values for YOLOv5 and YOLOv7 Fine-Tuning and GLIP Evaluation on 100 of Roboflow Universe
+datasets.
+
+
+## RF100
+
+`RF100` contains the following datasets, carefully chosen from more than 90'000 datasets hosted on our [universe hub](https://universe.roboflow.com/). The datasets are splitted in 7 categories: `Aerial`, `Videogames`, `Microscopic`, `Underwater`, `Documents`, `Electromagnetic` and `Real World`. 
+
+
 | Category        | Datasets | Images  | Classes |
 |-----------------|----------|---------|---------|
 | Aerial          | 7        | 9683    | 24      |
@@ -11,17 +21,10 @@
 | Documents       | 8        | 24813   | 90      |
 | Electromagnetic | 12       | 36381   | 41      |
 | Real World      | 50       | 110615  | 495     |
-| Total           | 100      | 224,714 | 805     |
+| **Total**           | **100**      | **224,714** | **805**     |
 
-This repository implements the Roboflow 100 benchmark developed by [Roboflow](https://roboflow.com/). It contains code to download the dataset and reproduce
-mAP values for YOLOv5 and YOLOv7 Fine-Tuning and GLIP Evaluation on 100 of Roboflow Universe
-datasets.
-
-## RF100
-
-`RF100` contains the following datasets, carefully chosen from more than 90'000 datasets hosted on our [universe hub](https://universe.roboflow.com/). The datasets are splitted in 7 categories: `Aerial`, `Videogames`, `Microscopic`, `Underwater`, `Documents`, `Electromagnetic` and `Real World`.
 
-**TODO** create a table
+🚨 **Head over the [Appendix](#appendix) at the bottom to see samples from each dataset**
 
 We provide a [notebook to help you using RF100 with PyTorch](/notebooks/roboflow-100-pytorch.ipynb)
 
@@ -154,3 +157,109 @@ nvidia-docker run --gpus all --rm -it --ipc host --network host --shm-size 64g \
     -v ${PWD}/datasets_links_640.txt:/workspace/datasets_links_640.txt \
     rf100-benchmark ./GLIP-benchmark/train.sh
 ```
+
+
+## Appendix
+
+| dataset                       | samples                                                   |
+|:------------------------------|:----------------------------------------------------------|
+| hand-gestures-jps7z           | ![alt](doc/images/grid/hand-gestures-jps7z.jpg)           |
+| smoke-uvylj                   | ![alt](doc/images/grid/smoke-uvylj.jpg)                   |
+| wall-damage                   | ![alt](doc/images/grid/wall-damage.jpg)                   |
+| corrosion-bi3q3               | ![alt](doc/images/grid/corrosion-bi3q3.jpg)               |
+| excavators-czvg9              | ![alt](doc/images/grid/excavators-czvg9.jpg)              |
+| chess-pieces-mjzgj            | ![alt](doc/images/grid/chess-pieces-mjzgj.jpg)            |
+| road-signs-6ih4y              | ![alt](doc/images/grid/road-signs-6ih4y.jpg)              |
+| street-work                   | ![alt](doc/images/grid/street-work.jpg)                   |
+| construction-safety-gsnvb     | ![alt](doc/images/grid/construction-safety-gsnvb.jpg)     |
+| road-traffic                  | ![alt](doc/images/grid/road-traffic.jpg)                  |
+| washroom-rf1fa                | ![alt](doc/images/grid/washroom-rf1fa.jpg)                |
+| circuit-elements              | ![alt](doc/images/grid/circuit-elements.jpg)              |
+| mask-wearing-608pr            | ![alt](doc/images/grid/mask-wearing-608pr.jpg)            |
+| cables-nl42k                  | ![alt](doc/images/grid/cables-nl42k.jpg)                  |
+| soda-bottles                  | ![alt](doc/images/grid/soda-bottles.jpg)                  |
+| truck-movement                | ![alt](doc/images/grid/truck-movement.jpg)                |
+| wine-labels                   | ![alt](doc/images/grid/wine-labels.jpg)                   |
+| digits-t2eg6                  | ![alt](doc/images/grid/digits-t2eg6.jpg)                  |
+| vehicles-q0x2v                | ![alt](doc/images/grid/vehicles-q0x2v.jpg)                |
+| peanuts-sd4kf                 | ![alt](doc/images/grid/peanuts-sd4kf.jpg)                 |
+| printed-circuit-board         | ![alt](doc/images/grid/printed-circuit-board.jpg)         |
+| pests-2xlvx                   | ![alt](doc/images/grid/pests-2xlvx.jpg)                   |
+| cavity-rs0uf                  | ![alt](doc/images/grid/cavity-rs0uf.jpg)                  |
+| leaf-disease-nsdsr            | ![alt](doc/images/grid/leaf-disease-nsdsr.jpg)            |
+| marbles                       | ![alt](doc/images/grid/marbles.jpg)                       |
+| pills-sxdht                   | ![alt](doc/images/grid/pills-sxdht.jpg)                   |
+| poker-cards-cxcvz             | ![alt](doc/images/grid/poker-cards-cxcvz.jpg)             |
+| number-ops                    | ![alt](doc/images/grid/number-ops.jpg)                    |
+| insects-mytwu                 | ![alt](doc/images/grid/insects-mytwu.jpg)                 |
+| cotton-20xz5                  | ![alt](doc/images/grid/cotton-20xz5.jpg)                  |
+| furniture-ngpea               | ![alt](doc/images/grid/furniture-ngpea.jpg)               |
+| cable-damage                  | ![alt](doc/images/grid/cable-damage.jpg)                  |
+| animals-ij5d2                 | ![alt](doc/images/grid/animals-ij5d2.jpg)                 |
+| coins-1apki                   | ![alt](doc/images/grid/coins-1apki.jpg)                   |
+| apples-fvpl5                  | ![alt](doc/images/grid/apples-fvpl5.jpg)                  |
+| people-in-paintings           | ![alt](doc/images/grid/people-in-paintings.jpg)           |
+| circuit-voltages              | ![alt](doc/images/grid/circuit-voltages.jpg)              |
+| uno-deck                      | ![alt](doc/images/grid/uno-deck.jpg)                      |
+| grass-weeds                   | ![alt](doc/images/grid/grass-weeds.jpg)                   |
+| gauge-u2lwv                   | ![alt](doc/images/grid/gauge-u2lwv.jpg)                   |
+| sign-language-sokdr           | ![alt](doc/images/grid/sign-language-sokdr.jpg)           |
+| valentines-chocolate          | ![alt](doc/images/grid/valentines-chocolate.jpg)          |
+| fish-market-ggjso             | ![alt](doc/images/grid/fish-market-ggjso.jpg)             |
+| lettuce-pallets               | ![alt](doc/images/grid/lettuce-pallets.jpg)               |
+| shark-teeth-5atku             | ![alt](doc/images/grid/shark-teeth-5atku.jpg)             |
+| bees-jt5in                    | ![alt](doc/images/grid/bees-jt5in.jpg)                    |
+| sedimentary-features-9eosf    | ![alt](doc/images/grid/sedimentary-features-9eosf.jpg)    |
+| currency-v4f8j                | ![alt](doc/images/grid/currency-v4f8j.jpg)                |
+| trail-camera                  | ![alt](doc/images/grid/trail-camera.jpg)                  |
+| cell-towers                   | ![alt](doc/images/grid/cell-towers.jpg)                   |
+| apex-videogame                | ![alt](doc/images/grid/apex-videogame.jpg)                |
+| farcry6-videogame             | ![alt](doc/images/grid/farcry6-videogame.jpg)             |
+| csgo-videogame                | ![alt](doc/images/grid/csgo-videogame.jpg)                |
+| avatar-recognition-nuexe      | ![alt](doc/images/grid/avatar-recognition-nuexe.jpg)      |
+| halo-infinite-angel-videogame | ![alt](doc/images/grid/halo-infinite-angel-videogame.jpg) |
+| team-fight-tactics            | ![alt](doc/images/grid/team-fight-tactics.jpg)            |
+| robomasters-285km             | ![alt](doc/images/grid/robomasters-285km.jpg)             |
+| tweeter-posts                 | ![alt](doc/images/grid/tweeter-posts.jpg)                 |
+| tweeter-profile               | ![alt](doc/images/grid/tweeter-profile.jpg)               |
+| document-parts                | ![alt](doc/images/grid/document-parts.jpg)                |
+| activity-diagrams-qdobr       | ![alt](doc/images/grid/activity-diagrams-qdobr.jpg)       |
+| signatures-xc8up              | ![alt](doc/images/grid/signatures-xc8up.jpg)              |
+| paper-parts                   | ![alt](doc/images/grid/paper-parts.jpg)                   |
+| tabular-data-wf9uh            | ![alt](doc/images/grid/tabular-data-wf9uh.jpg)            |
+| paragraphs-co84b              | ![alt](doc/images/grid/paragraphs-co84b.jpg)              |
+| underwater-pipes-4ng4t        | ![alt](doc/images/grid/underwater-pipes-4ng4t.jpg)        |
+| aquarium-qlnqy                | ![alt](doc/images/grid/aquarium-qlnqy.jpg)                |
+| peixos-fish                   | ![alt](doc/images/grid/peixos-fish.jpg)                   |
+| underwater-objects-5v7p8      | ![alt](doc/images/grid/underwater-objects-5v7p8.jpg)      |
+| coral-lwptl                   | ![alt](doc/images/grid/coral-lwptl.jpg)                   |
+| aerial-pool                   | ![alt](doc/images/grid/aerial-pool.jpg)                   |
+| secondary-chains              | ![alt](doc/images/grid/secondary-chains.jpg)              |
+| aerial-spheres                | ![alt](doc/images/grid/aerial-spheres.jpg)                |
+| soccer-players-5fuqs          | ![alt](doc/images/grid/soccer-players-5fuqs.jpg)          |
+| weed-crop-aerial              | ![alt](doc/images/grid/weed-crop-aerial.jpg)              |
+| aerial-cows                   | ![alt](doc/images/grid/aerial-cows.jpg)                   |
+| cloud-types                   | ![alt](doc/images/grid/cloud-types.jpg)                   |
+| stomata-cells                 | ![alt](doc/images/grid/stomata-cells.jpg)                 |
+| bccd-ouzjz                    | ![alt](doc/images/grid/bccd-ouzjz.jpg)                    |
+| parasites-1s07h               | ![alt](doc/images/grid/parasites-1s07h.jpg)               |
+| cells-uyemf                   | ![alt](doc/images/grid/cells-uyemf.jpg)                   |
+| 4-fold-defect                 | ![alt](doc/images/grid/4-fold-defect.jpg)                 |
+| bacteria-ptywi                | ![alt](doc/images/grid/bacteria-ptywi.jpg)                |
+| cotton-plant-disease          | ![alt](doc/images/grid/cotton-plant-disease.jpg)          |
+| mitosis-gjs3g                 | ![alt](doc/images/grid/mitosis-gjs3g.jpg)                 |
+| phages                        | ![alt](doc/images/grid/phages.jpg)                        |
+| liver-disease                 | ![alt](doc/images/grid/liver-disease.jpg)                 |
+| asbestos                      | ![alt](doc/images/grid/asbestos.jpg)                      |
+| thermal-dogs-and-people-x6ejw | ![alt](doc/images/grid/thermal-dogs-and-people-x6ejw.jpg) |
+| solar-panels-taxvb            | ![alt](doc/images/grid/solar-panels-taxvb.jpg)            |
+| radio-signal                  | ![alt](doc/images/grid/radio-signal.jpg)                  |
+| thermal-cheetah-my4dp         | ![alt](doc/images/grid/thermal-cheetah-my4dp.jpg)         |
+| x-ray-rheumatology            | ![alt](doc/images/grid/x-ray-rheumatology.jpg)            |
+| acl-x-ray                     | ![alt](doc/images/grid/acl-x-ray.jpg)                     |
+| abdomen-mri                   | ![alt](doc/images/grid/abdomen-mri.jpg)                   |
+| axial-mri                     | ![alt](doc/images/grid/axial-mri.jpg)                     |
+| gynecology-mri                | ![alt](doc/images/grid/gynecology-mri.jpg)                |
+| brain-tumor-m2pbp             | ![alt](doc/images/grid/brain-tumor-m2pbp.jpg)             |
+| bone-fracture-7fylg           | ![alt](doc/images/grid/bone-fracture-7fylg.jpg)           |
+| flir-camera-objects           | ![alt](doc/images/grid/flir-camera-objects.jpg)           |

scripts/convert_table_from_old_flo_paper_to_csv.py

@@ -1,8 +1,9 @@
 """
 Poor man script to convert the table from an old version of the paper. This was the only way since we didn't have access to the original data source from the table
 """
-from pathlib import Path
 from argparse import ArgumentParser
+from pathlib import Path
+
 import pandas as pd
 
 parser = ArgumentParser()

scripts/download_dataset.py

@@ -1,6 +1,7 @@
 from argparse import ArgumentParser
 from os import environ
 from pathlib import Path
+
 from roboflow import Roboflow
 
 

scripts/get_labels_names.py

@@ -1,9 +1,10 @@
 import os
-import requests
 from concurrent.futures import ThreadPoolExecutor
+from pprint import pprint
+
 import pandas as pd
+import requests
 from tqdm import tqdm
-from pprint import pprint
 
 ROBOFLOW_API_KEY = os.environ["ROBOFLOW_API_KEY"]
 API_URL = "https://api.roboflow.com/roboflow-100"

scripts/image/cluster.py

@@ -1,16 +1,16 @@
-import numpy as np
-from pickle import load, dump
+from multiprocessing import Pool
 from pathlib import Path
-from tqdm import tqdm
+from pickle import dump, load
+
+import numpy as np
+import pandas as pd
 import torch
-from torchvision.utils import make_grid
 import torchvision.transforms as T
-from torchvision.transforms.functional import to_pil_image
-import pandas as pd
-from PIL import Image
-from multiprocessing import Pool
-
 from kmeans import kmeans
+from PIL import Image
+from torchvision.transforms.functional import to_pil_image
+from torchvision.utils import make_grid
+from tqdm import tqdm
 
 NUM_CLUSTERS = 32 * 56
 

scripts/image/encode.py

@@ -1,17 +1,17 @@
 # read images from rf100
 # encode them with clip
 
+from pathlib import Path
+from pickle import dump
 from typing import Callable, Optional
-import torch
+
 import clip
-from PIL import Image
-from pathlib import Path
+import numpy as np
+import torch
 import torch as nn
-from torch.utils.data import Dataset, DataLoader
 from PIL import Image
-import numpy as np
+from torch.utils.data import DataLoader, Dataset
 from torchvision.transforms import ToTensor
-from pickle import dump
 from tqdm import tqdm
 
 device = "cuda" if torch.cuda.is_available() else "cpu"

scripts/image/kmeans.py

@@ -1,6 +1,6 @@
 import torch
-from einops import rearrange, repeat
 import torch.nn.functional as F
+from einops import rearrange, repeat
 from tqdm import tqdm
 
 

scripts/make_dataset_table_with_images.py

@@ -0,0 +1,57 @@
+"""Code to generate the markdown table in the README.md
+"""
+import random
+from pathlib import Path
+from typing import Tuple
+
+import pandas as pd
+import torch
+from PIL import Image
+from torchvision.transforms.functional import pil_to_tensor, to_pil_image
+from torchvision.utils import make_grid
+
+ROOT = Path("./rf100")
+
+
+def read_and_transform(image_path: Path) -> torch.Tensor:
+    image = Image.open(image_path).convert("RGB").resize((128, 128))
+    return pil_to_tensor(image)
+
+
+def make_grid_from_dataset(dataset_dir: Path, n=5) -> Tuple[str, Image.Image]:
+    # get random 5 images
+    images_paths = list((dataset_dir / "train" / "images").glob("*"))
+    sampled_images_paths = random.choices(images_paths, k=n)
+    tensors = [read_and_transform(image_path) for image_path in sampled_images_paths]
+    grid = make_grid(tensors, nrow=n)
+    return dataset_dir.stem, to_pil_image(grid)
+
+
+def save(dataset_name: str, grid: Image.Image) -> Tuple[str, str]:
+    grid_path = f"doc/images/grid/{dataset_name}.jpg"
+    # grid.save(grid_path, quality=70)
+    return dataset_name, grid_path
+
+
+def make_grids(root: Path):
+    out = list(map(make_grid_from_dataset, root.iterdir()))
+    out = list(map(lambda x: save(*x), out))
+    return out
+
+
+def make_table(root: Path) -> str:
+    out = make_grids(root)
+    dataset_names, grid_paths = zip(*out)
+    category_df = pd.read_csv("metadata/categories.csv", index_col=0)
+    df = pd.DataFrame(
+        data=dict(dataset=dataset_names, samples=grid_paths),
+        columns=["dataset", "samples"],
+    )
+    df.samples = df.samples.apply(lambda x: f"![alt]({x})")
+    df = df.set_index("dataset", drop=True)
+    df = category_df.join(df)
+    del df["category"]
+    return df.to_markdown()
+
+
+print(make_table(ROOT))

scripts/make_results_table.py

@@ -9,6 +9,7 @@
 from pathlib import Path
 from textwrap import indent
 from unicodedata import name
+
 import pandas as pd
 
 RF100_BASE_URL = "https://app.roboflow.com/roboflow-100"

scripts/parse_categories.py

@@ -1,10 +1,11 @@
 """
 This script takes a file text (lost in time) and outputs a correctly formated csv file with category -> dataset
 """
+import re
 from argparse import ArgumentParser
 from pathlib import Path
+
 import pandas as pd
-import re
 
 parser = ArgumentParser()
 parser.add_argument("-i", "--input", type=Path)

scripts/reduce_dims.py

@@ -1,12 +1,12 @@
-from typing import Dict, Iterator, Tuple, List
-from image.pca import pca
 from pathlib import Path
-from tqdm import tqdm
-from pickle import load, dump
+from pickle import dump, load
+from typing import Dict, Iterator, List, Tuple
+
 import numpy as np
 import torch
-
+from image.pca import pca
 from sklearn.manifold import TSNE
+from tqdm import tqdm
 
 
 def get_data(root: Path) -> dict:

scripts/split_for_webapp.py

@@ -2,14 +2,15 @@
 This script will split the reduced-* pickle file into folders with the right dimension and metadata for the collage.
 """
 
+import json
+from concurrent.futures import ThreadPoolExecutor
 from pathlib import Path
-from typing import Iterable, Tuple, List
-from tqdm import tqdm
-from pickle import load, dump
+from pickle import dump, load
+from typing import Iterable, List, Tuple
+
 import numpy as np
-from concurrent.futures import ThreadPoolExecutor
 from PIL import Image
-import json
+from tqdm import tqdm
 
 """
 1) read the data
@@ -90,7 +91,7 @@ def get_data_chunks(
 
 def make_split(
     data: dict, output_dir: Path, montage_size: str, image_size: str, *args, **kwargs
-):  
+):
     """
     Takes `data`, writes to disk all the images and the embeddings in json format
     """
@@ -113,7 +114,6 @@ def make_split(
         )
 
 
-
 def make_splits_by_dataset(
     data: dict, output_dir: Path, montage_size: str, image_size: str, **kwargs
 ) -> List[dict]: