Added modification based on customized dataset

datasets/ucsd.py

@@ -0,0 +1,27 @@
+"""
+For UCSD dataset.
+
+"""
+from pathlib import Path
+
+from .coco import CocoDetection, make_coco_transforms
+
+def build(image_set, args):
+    """
+    image_set = 'train' / 'val'
+    category = "all/tflt/"
+    """
+    root = Path(args.coco_path)
+    assert root.exists(), f'provided path {root} to custom dataset does not exist'
+
+    if args.annotation_name is None:
+        raise ValueError("args doesn't have annotation_name")
+    else:
+        PATHS = {
+            "train": (root / "train2017", root / "annotations" / f'annotations_train_{args.annotation_name}.json'),
+            "val": (root / "val2017", root / "annotations" / f'annotations_val_{args.annotation_name}.json'),
+        }
+
+    img_folder, ann_file = PATHS[image_set]
+    dataset = CocoDetection(img_folder, ann_file, transforms=make_coco_transforms(image_set), return_masks=args.masks)
+    return dataset

main.py

@@ -12,19 +12,29 @@
 
 import datasets
 import util.misc as utils
-from datasets import build_dataset, get_coco_api_from_dataset
+from datasets import get_coco_api_from_dataset
+from datasets.ucsd import build as build_dataset
 from engine import evaluate, train_one_epoch
 from models import build_model
 
 
 def get_args_parser():
     parser = argparse.ArgumentParser('Set transformer detector', add_help=False)
+    # added for UCSD
+    parser.add_argument('--num_classes', default=None, type=int,
+                        help='num of classes in your dataset, which can override the value hard-coded in file models/detr.py')
+    parser.add_argument('--annotation_name', default=None, type=str, action='store_true',
+                        help='all / tfsg / veh / tflt')
+    parser.add_argument('--gpu_id', default=None, type=int,
+                        help="specify GPU ID")
+
+    # original
     parser.add_argument('--lr', default=1e-4, type=float)
     parser.add_argument('--lr_backbone', default=1e-5, type=float)
     parser.add_argument('--batch_size', default=2, type=int)
     parser.add_argument('--weight_decay', default=1e-4, type=float)
     parser.add_argument('--epochs', default=300, type=int)
-    parser.add_argument('--lr_drop', default=200, type=int)
+    parser.add_argument('--lr_drop', default=100, type=int)
     parser.add_argument('--clip_max_norm', default=0.1, type=float,
                         help='gradient clipping max norm')
 
@@ -111,6 +121,7 @@ def main(args):
     print(args)
 
     device = torch.device(args.device)
+    torch.cuda.set_device(args.gpu_id)
 
     # fix the seed for reproducibility
     seed = args.seed + utils.get_rank()
@@ -121,10 +132,14 @@ def main(args):
     model, criterion, postprocessors = build_model(args)
     model.to(device)
 
-    model_without_ddp = model
     if args.distributed:
         model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu])
         model_without_ddp = model.module
+    else:
+        print("Data parallel mode activated!")
+        model = torch.nn.DataParallel(model, device_ids=[args.gpu_id])
+        model = model.to(device)
+        model_without_ddp = model
     n_parameters = sum(p.numel() for p in model.parameters() if p.requires_grad)
     print('number of params:', n_parameters)
 
@@ -200,11 +215,11 @@ def main(args):
         if args.output_dir:
             checkpoint_paths = [output_dir / 'checkpoint.pth']
             # extra checkpoint before LR drop and every 100 epochs
-            if (epoch + 1) % args.lr_drop == 0 or (epoch + 1) % 100 == 0:
+            if (epoch + 1) % args.lr_drop == 0 or (epoch + 1) % 10 == 0:
                 checkpoint_paths.append(output_dir / f'checkpoint{epoch:04}.pth')
             for checkpoint_path in checkpoint_paths:
                 utils.save_on_master({
-                    'model': model_without_ddp.state_dict(),
+                    'model': model.state_dict(),
                     'optimizer': optimizer.state_dict(),
                     'lr_scheduler': lr_scheduler.state_dict(),
                     'epoch': epoch,
@@ -229,7 +244,7 @@ def main(args):
                 (output_dir / 'eval').mkdir(exist_ok=True)
                 if "bbox" in coco_evaluator.coco_eval:
                     filenames = ['latest.pth']
-                    if epoch % 50 == 0:
+                    if epoch % 10 == 0:
                         filenames.append(f'{epoch:03}.pth')
                     for name in filenames:
                         torch.save(coco_evaluator.coco_eval["bbox"].eval,
@@ -246,3 +261,5 @@ def main(args):
     if args.output_dir:
         Path(args.output_dir).mkdir(parents=True, exist_ok=True)
     main(args)
+
+## python -m torch.distributed.launch --nproc_per_node=8 --use_env main.py --coco_path /path/to/coco

models/detr.py

@@ -302,14 +302,6 @@ def forward(self, x):
 
 
 def build(args):
-    # the `num_classes` naming here is somewhat misleading.
-    # it indeed corresponds to `max_obj_id + 1`, where max_obj_id
-    # is the maximum id for a class in your dataset. For example,
-    # COCO has a max_obj_id of 90, so we pass `num_classes` to be 91.
-    # As another example, for a dataset that has a single class with id 1,
-    # you should pass `num_classes` to be 2 (max_obj_id + 1).
-    # For more details on this, check the following discussion
-    # https://github.com/facebookresearch/detr/issues/108#issuecomment-650269223
     num_classes = 20 if args.dataset_file != 'coco' else 91
     if args.dataset_file == "coco_panoptic":
         # for panoptic, we just add a num_classes that is large enough to hold