dataloader modified

data/hyps/hyp.scratch-high.yaml

@@ -26,5 +26,5 @@ perspective: 0.0  # image perspective (+/- fraction), range 0-0.001
 flipud: 0.0  # image flip up-down (probability)
 fliplr: 0.5  # image flip left-right (probability)
 mosaic: 1.0  # image mosaic (probability)
-mixup: 0.15  # image mixup (probability)
-copy_paste: 0.3  # segment copy-paste (probability)
+mixup: 0.5  # image mixup (probability)
+copy_paste: 0.0  # segment copy-paste (probability)

data/polyp_2.yaml

@@ -0,0 +1,18 @@
+# COCO 2017 dataset http://cocodataset.org
+
+# download command/URL (optional)
+# download: bash ./scripts/get_coco.sh
+
+# train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/]
+# train: ./coco/train2017.txt  # 118287 images
+# val: ./coco/val2017.txt  # 5000 images
+# test: ./coco/test-dev2017.txt  # 20288 of 40670 images, submit to https://competitions.codalab.org/competitions/20794
+train: /work/u6859530/polyp_detection/dataset2/images/train
+val: /work/u6859530/polyp_detection/dataset2/images/valid
+test: /work/u6859530/polyp_detection/dataset/images/valid
+
+# number of classes
+nc: 2
+
+# class names
+names: [ 'hyperplastic', 'adenoma']

models/common.py

@@ -413,7 +413,7 @@ def forward(self, x):
             warnings.simplefilter('ignore')  # suppress torch 1.9.0 max_pool2d() warning
             return self.cv2(torch.cat([x] + [m(x) for m in self.m], 1))
 
-        
+
 class ASPP(torch.nn.Module):
 
     def __init__(self, in_channels, out_channels):
@@ -493,8 +493,8 @@ def forward(self, x):
 
 import torch.nn.functional as F
 from torch.nn.modules.utils import _pair
-    
-    
+
+
 class ReOrg(nn.Module):
     # yolo
     def __init__(self):
@@ -549,17 +549,17 @@ def __init__(self, dimension=0):
 
     def forward(self, x):
         return x[0]+x[1]
-    
-    
+
+
 class Silence(nn.Module):
     def __init__(self):
         super(Silence, self).__init__()
     def forward(self, x):    
         return x
 
 
-##### GELAN #####        
-        
+# #### GELAN #####        
+
 class SPPELAN(nn.Module):
     # spp-elan
     def __init__(self, c1, c2, c3):  # ch_in, ch_out, number, shortcut, groups, expansion
@@ -575,8 +575,8 @@ def forward(self, x):
         y = [self.cv1(x)]
         y.extend(m(y[-1]) for m in [self.cv2, self.cv3, self.cv4])
         return self.cv5(torch.cat(y, 1))
-        
-        
+
+
 class RepNCSPELAN4(nn.Module):
     # csp-elan
     def __init__(self, c1, c2, c3, c4, c5=1):  # ch_in, ch_out, number, shortcut, groups, expansion
@@ -597,10 +597,10 @@ def forward_split(self, x):
         y.extend(m(y[-1]) for m in [self.cv2, self.cv3])
         return self.cv4(torch.cat(y, 1))
 
-#################
+# ################
 
 
-##### YOLOR #####
+# #### YOLOR #####
 
 class ImplicitA(nn.Module):
     def __init__(self, channel):
@@ -623,10 +623,10 @@ def __init__(self, channel):
     def forward(self, x):
         return self.implicit * x
 
-#################
+# ################
 
 
-##### CBNet #####
+# #### CBNet #####
 
 class CBLinear(nn.Module):
     def __init__(self, c1, c2s, k=1, s=1, p=None, g=1):  # ch_in, ch_outs, kernel, stride, padding, groups
@@ -649,7 +649,7 @@ def forward(self, xs):
         out = torch.sum(torch.stack(res + xs[-1:]), dim=0)
         return out
 
-#################
+# ################
 
 
 class DetectMultiBackend(nn.Module):

models/yolo.py

@@ -464,7 +464,7 @@ def forward(self, x):
         if self.training:
             return x, mc, p, s
         return (torch.cat([x, mc], 1), p, s) if self.export else (torch.cat([x[0], mc], 1), (x[1], mc, p, s))
-    
+
 
 class BaseModel(nn.Module):
     # YOLO base model

requirements.txt

@@ -2,8 +2,6 @@
 # Usage: pip install -r requirements.txt
 
 # Base ------------------------------------------------------------------------
-gitpython
-ipython
 matplotlib>=3.2.2
 numpy>=1.18.5
 opencv-python>=4.1.1

run.sh

@@ -0,0 +1,7 @@
+# train yolov9 models
+python train_dual.py --workers 8 --device 0 --batch 16 --data data/polyp_2.yaml --img 640 --cfg models/detect/yolov9-e.yaml --weights '' --name v9-e_1c --hyp hyp.scratch-high.yaml --min-items 0 --epochs 300 --close-mosaic 0 --exist-ok --single-cls
+python val_dual.py --data data/polyp_2.yaml --img 640 --batch 32 --conf 0.25 --iou 0.65 --device 0 --weights 'runs/v9-e_1c/weights/best.pt' --name v9-c_1c --exist-ok --single-cls
+python train_dual.py --workers 8 --device 0 --batch 16 --data data/polyp_2.yaml --img 640 --cfg models/detect/yolov9-e.yaml --weights '' --name v9-e_2c --hyp hyp.scratch-high.yaml --min-items 0 --epochs 300 --close-mosaic 0 --exist-ok
+python val_dual.py --data data/polyp_2.yaml --img 640 --batch 32 --conf 0.25 --iou 0.65 --device 0 --weights 'runs/v9-e_2c/weights/best.pt' --name v9-e_2c --exist-ok
+python train_dual.py --workers 8 --device 0 --batch 16 --data data/polyp_2.yaml --img 640 --cfg models/detect/yolov9-c.yaml --weights '' --name v9-c_1c --hyp hyp.scratch-high.yaml --min-items 0 --epochs 300 --close-mosaic 0 --exist-ok --single-cls
+python val_dual.py --data data/polyp_2.yaml --img 640 --batch 32 --conf 0.25 --iou 0.65 --device 0 --weights 'runs/v9-c_1c/weights/best.pt' --name v9-c_1c --exist-ok --single-cls

train_dual.py

@@ -56,7 +56,6 @@ def train(hyp, opt, device, callbacks):  # hyp is path/to/hyp.yaml or hyp dictio
         Path(opt.save_dir), opt.epochs, opt.batch_size, opt.weights, opt.single_cls, opt.evolve, opt.data, opt.cfg, \
         opt.resume, opt.noval, opt.nosave, opt.workers, opt.freeze
     callbacks.run('on_pretrain_routine_start')
-
     # Directories
     w = save_dir / 'weights'  # weights dir
     (w.parent if evolve else w).mkdir(parents=True, exist_ok=True)  # make dir
@@ -117,7 +116,6 @@ def train(hyp, opt, device, callbacks):  # hyp is path/to/hyp.yaml or hyp dictio
     else:
         model = Model(cfg, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device)  # create
     amp = check_amp(model)  # check AMP
-
     # Freeze
     freeze = [f'model.{x}.' for x in (freeze if len(freeze) > 1 else range(freeze[0]))]  # layers to freeze
     for k, v in model.named_parameters():
@@ -285,7 +283,11 @@ def train(hyp, opt, device, callbacks):  # hyp is path/to/hyp.yaml or hyp dictio
         if RANK in {-1, 0}:
             pbar = tqdm(pbar, total=nb, bar_format=TQDM_BAR_FORMAT)  # progress bar
         optimizer.zero_grad()
+        #tt0 = time.time()*1000
         for i, (imgs, targets, paths, _) in pbar:  # batch -------------------------------------------------------------
+            #tt1 = time.time()*1000
+            #print("\n loading time: ", tt1-tt0)
+            #tt0 = tt1
             callbacks.run('on_train_batch_start')
             ni = i + nb * epoch  # number integrated batches (since train start)
             imgs = imgs.to(device, non_blocking=True).float() / 255  # uint8 to float32, 0-255 to 0.0-1.0
@@ -341,6 +343,7 @@ def train(hyp, opt, device, callbacks):  # hyp is path/to/hyp.yaml or hyp dictio
                 callbacks.run('on_train_batch_end', model, ni, imgs, targets, paths, list(mloss))
                 if callbacks.stop_training:
                     return
+            #print("\nend: ",  time.time()*1000-t_start, "ms")
             # end batch ------------------------------------------------------------------------------------------------
 
         # Scheduler
@@ -462,7 +465,7 @@ def parse_opt(known=False):
     parser.add_argument('--optimizer', type=str, choices=['SGD', 'Adam', 'AdamW', 'LION'], default='SGD', help='optimizer')
     parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
     parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)')
-    parser.add_argument('--project', default=ROOT / 'runs/train', help='save to project/name')
+    parser.add_argument('--project', default=ROOT / 'runs/', help='save to project/name')
     parser.add_argument('--name', default='exp', help='save to project/name')
     parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
     parser.add_argument('--quad', action='store_true', help='quad dataloader')