chore: remove unnecessary things from YOLOPv2 demo

demos/yolopv2/requirements.txt

@@ -1,30 +1,6 @@
 # Usage: pip install -r requirements.txt
 
-# Base ----------------------------------------
-matplotlib>=3.2.2
-numpy>=1.18.5
 opencv-python==4.5.5.64
-Pillow>=7.1.2
-PyYAML>=5.3.1
-requests>=2.23.0
-scipy>=1.4.1
-tqdm>=4.41.0
-protobuf<4.21.3
-
-# Torch
---find-links https://download.pytorch.org/whl/cu116
-torch==1.12.0+cu116 
-torchvision==0.13.0+cu116 
-torchaudio==0.12.0
-
-# Logging -------------------------------------
-tensorboard>=2.4.1
-
-# Plotting ------------------------------------
-pandas>=1.1.4
-seaborn>=0.11.0
-
-# Extras --------------------------------------
-ipython  # interactive notebook
-psutil  # system utilization
-thop  # FLOPs computation
+torch==1.13.0
+torchvision==1.13.0
+numpy==1.22.4

demos/yolopv2/src/demo.py

@@ -8,18 +8,13 @@
 
 # Conclude setting / general reprocessing / plots / metrices / datasets
 from utils import (
-    AverageMeter,
     LoadImages,
     driving_area_mask,
     increment_path,
     lane_line_mask,
     non_max_suppression,
-    plot_one_box,
     scale_coords,
-    select_device,
-    show_seg_result,
     split_for_trace_model,
-    time_synchronized,
     yolop_detections_to_norfair_detections,
 )
 
@@ -93,15 +88,10 @@ def detect():
         parents=True, exist_ok=True
     )  # make dir
 
-    inf_time = AverageMeter()
-    waste_time = AverageMeter()
-    nms_time = AverageMeter()
-
     # Load model
-    stride = 32
     model = torch.jit.load(weights)
-    device = select_device(opt.device)
-    half = device.type != "cpu"  # half precision only supported on CUDA
+    device = "cuda:0" if torch.cuda.is_available() else "cpu"
+    half = device != "cpu"  # half precision only supported on CUDA
     model = model.to(device)
 
     # Norfair Tracker init
@@ -116,10 +106,10 @@ def detect():
 
     # Set Dataloader
     vid_path, vid_writer = None, None
-    dataset = LoadImages(source, img_size=imgsz, stride=stride)
+    dataset = LoadImages(source, img_size=imgsz)
 
     # Run inference
-    if device.type != "cpu":
+    if device != "cpu":
         model(
             torch.zeros(1, 3, imgsz, imgsz).to(device).type_as(next(model.parameters()))
         )  # run once
@@ -134,26 +124,18 @@ def detect():
             img = img.unsqueeze(0)
 
         # Inference
-        t1 = time_synchronized()
         [pred, anchor_grid], seg, ll = model(img)
-        t2 = time_synchronized()
 
-        # waste time: the incompatibility of  torch.jit.trace causes extra time consumption in demo version
-        # but this problem will not appear in offical version
-        tw1 = time_synchronized()
         pred = split_for_trace_model(pred, anchor_grid)
-        tw2 = time_synchronized()
 
         # Apply NMS
-        t3 = time_synchronized()
         pred = non_max_suppression(
             pred,
             opt.conf_thres,
             opt.iou_thres,
             classes=opt.classes,
             agnostic=opt.agnostic_nms,
         )
-        t4 = time_synchronized()
 
         da_seg_mask = driving_area_mask(seg)
         ll_seg_mask = lane_line_mask(ll)
@@ -177,10 +159,6 @@ def detect():
             save_path = str(save_dir / p.name)  # img.jpg
             s += "%gx%g " % img.shape[2:]  # print string
 
-            # Print time (inference)
-            print(f"{s}Done. ({t2 - t1:.3f}s)")
-            show_seg_result(im0, (da_seg_mask, ll_seg_mask), is_demo=True)
-
             # Save results (image with detections)
             if save_img:
                 if dataset.mode == "image":
@@ -202,11 +180,7 @@ def detect():
                         )
                     vid_writer.write(im0)
 
-    inf_time.update(t2 - t1, img.size(0))
-    nms_time.update(t4 - t3, img.size(0))
-    waste_time.update(tw2 - tw1, img.size(0))
-    print("inf : (%.4fs/frame)   nms : (%.4fs/frame)" % (inf_time.avg, nms_time.avg))
-    print(f"Done. ({time.time() - t0:.3f}s)")
+    print(f"\nDone. ({time.time() - t0:.3f}s)")
 
 
 if __name__ == "__main__":