FP16 inference update

detect.py

@@ -14,6 +14,7 @@ def detect(save_img=False):
     if os.path.exists(out):
         shutil.rmtree(out)  # delete output folder
     os.makedirs(out)  # make new output folder
+    half &= device.type != 'cpu'  # half precision only supported on CUDA
 
     # Load model
     google_utils.attempt_download(weights)
@@ -21,6 +22,8 @@ def detect(save_img=False):
     # torch.save(torch.load(weights, map_location=device), weights)  # update model if SourceChangeWarning
     # model.fuse()
     model.to(device).eval()
+    if half:
+        model.half()  # to FP16
 
     # Second-stage classifier
     classify = False
@@ -29,11 +32,6 @@ def detect(save_img=False):
         modelc.load_state_dict(torch.load('weights/resnet101.pt', map_location=device)['model'])  # load weights
         modelc.to(device).eval()
 
-    # Half precision
-    half = half and device.type != 'cpu'  # half precision only supported on CUDA
-    if half:
-        model.half()
-
     # Set Dataloader
     vid_path, vid_writer = None, None
     if webcam:
@@ -51,7 +49,7 @@ def detect(save_img=False):
     # Run inference
     t0 = time.time()
     img = torch.zeros((1, 3, imgsz, imgsz), device=device)  # init img
-    _ = model(img.half() if half else img.float()) if device.type != 'cpu' else None  # run once
+    _ = model(img.half() if half else img) if device.type != 'cpu' else None  # run once
     for path, img, im0s, vid_cap in dataset:
         img = torch.from_numpy(img).to(device)
         img = img.half() if half else img.float()  # uint8 to fp16/32
@@ -63,10 +61,6 @@ def detect(save_img=False):
         t1 = torch_utils.time_synchronized()
         pred = model(img, augment=opt.augment)[0]
 
-        # to float
-        if half:
-            pred = pred.float()
-
         # Apply NMS
         pred = non_max_suppression(pred, opt.conf_thres, opt.iou_thres,
                                    fast=True, classes=opt.classes, agnostic=opt.agnostic_nms)

requirements.txt

 # pip install -U -r requirements.txt
 Cython
-numpy
+numpy==1.17
 opencv-python
-torch >= 1.5
+torch>=1.5
 matplotlib
 pillow
 tensorboard
-pyyaml >= 5.3
+PyYAML>=5.3
 torchvision
 scipy
 tqdm

test.py

@@ -20,10 +20,12 @@ def test(data,
          model=None,
          dataloader=None,
          fast=False,
-         verbose=False):  # 0 fast, 1 accurate
+         verbose=False,
+         half=False):  # FP16
     # Initialize/load model and set device
     if model is None:
         device = torch_utils.select_device(opt.device, batch_size=batch_size)
+        half &= device.type != 'cpu'  # half precision only supported on CUDA
 
         # Remove previous
         for f in glob.glob('test_batch*.jpg'):
@@ -35,6 +37,8 @@ def test(data,
         torch_utils.model_info(model)
         # model.fuse()
         model.to(device)
+        if half:
+            model.half()  # to FP16
 
         if device.type != 'cpu' and torch.cuda.device_count() > 1:
             model = nn.DataParallel(model)
@@ -72,24 +76,27 @@ def test(data,
 
     seen = 0
     model.eval()
-    _ = model(torch.zeros((1, 3, imgsz, imgsz), device=device)) if device.type != 'cpu' else None  # run once
+    img = torch.zeros((1, 3, imgsz, imgsz), device=device)  # init img
+    _ = model(img.half() if half else img) if device.type != 'cpu' else None  # run once
     names = model.names if hasattr(model, 'names') else model.module.names
     coco91class = coco80_to_coco91_class()
     s = ('%20s' + '%12s' * 6) % ('Class', 'Images', 'Targets', 'P', 'R', 'mAP@.5', 'mAP@.5:.95')
     p, r, f1, mp, mr, map50, map, t0, t1 = 0., 0., 0., 0., 0., 0., 0., 0., 0.
     loss = torch.zeros(3, device=device)
     jdict, stats, ap, ap_class = [], [], [], []
-    for batch_i, (imgs, targets, paths, shapes) in enumerate(tqdm(dataloader, desc=s)):
-        imgs = imgs.to(device).float() / 255.0  # uint8 to float32, 0 - 255 to 0.0 - 1.0
+    for batch_i, (img, targets, paths, shapes) in enumerate(tqdm(dataloader, desc=s)):
+        img = img.to(device)
+        img = img.half() if half else img.float()  # uint8 to fp16/32
+        img /= 255.0  # 0 - 255 to 0.0 - 1.0
         targets = targets.to(device)
-        nb, _, height, width = imgs.shape  # batch size, channels, height, width
+        nb, _, height, width = img.shape  # batch size, channels, height, width
         whwh = torch.Tensor([width, height, width, height]).to(device)
 
         # Disable gradients
         with torch.no_grad():
             # Run model
             t = torch_utils.time_synchronized()
-            inf_out, train_out = model(imgs, augment=augment)  # inference and training outputs
+            inf_out, train_out = model(img, augment=augment)  # inference and training outputs
             t0 += torch_utils.time_synchronized() - t
 
             # Compute loss
@@ -125,7 +132,7 @@ def test(data,
                 # [{"image_id": 42, "category_id": 18, "bbox": [258.15, 41.29, 348.26, 243.78], "score": 0.236}, ...
                 image_id = int(Path(paths[si]).stem.split('_')[-1])
                 box = pred[:, :4].clone()  # xyxy
-                scale_coords(imgs[si].shape[1:], box, shapes[si][0], shapes[si][1])  # to original shape
+                scale_coords(img[si].shape[1:], box, shapes[si][0], shapes[si][1])  # to original shape
                 box = xyxy2xywh(box)  # xywh
                 box[:, :2] -= box[:, 2:] / 2  # xy center to top-left corner
                 for p, b in zip(pred.tolist(), box.tolist()):
@@ -168,9 +175,9 @@ def test(data,
         # Plot images
         if batch_i < 1:
             f = 'test_batch%g_gt.jpg' % batch_i  # filename
-            plot_images(imgs, targets, paths, f, names)  # ground truth
+            plot_images(img, targets, paths, f, names)  # ground truth
             f = 'test_batch%g_pred.jpg' % batch_i
-            plot_images(imgs, output_to_target(output, width, height), paths, f, names)  # predictions
+            plot_images(img, output_to_target(output, width, height), paths, f, names)  # predictions
 
     # Compute statistics
     stats = [np.concatenate(x, 0) for x in zip(*stats)]  # to numpy
@@ -241,6 +248,7 @@ if __name__ == '__main__':
     parser.add_argument('--save-json', action='store_true', help='save a cocoapi-compatible JSON results file')
     parser.add_argument('--task', default='val', help="'val', 'test', 'study'")
     parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--half', action='store_true', help='half precision FP16 inference')
     parser.add_argument('--single-cls', action='store_true', help='treat as single-class dataset')
     parser.add_argument('--augment', action='store_true', help='augmented inference')
     parser.add_argument('--verbose', action='store_true', help='report mAP by class')

utils/utils.py

@@ -504,6 +504,9 @@ def non_max_suppression(prediction, conf_thres=0.1, iou_thres=0.6, fast=False, c
     Returns detections with shape:
         nx6 (x1, y1, x2, y2, conf, cls)
     """
+    if prediction.dtype is torch.float16:
+        prediction = prediction.float()  # to FP32
+
     nc = prediction[0].shape[1] - 5  # number of classes
     xc = prediction[..., 4] > conf_thres  # candidates
 
@@ -902,7 +905,7 @@ def plot_images(images, targets, paths=None, fname='images.jpg', names=None, max
         return None
 
     if isinstance(images, torch.Tensor):
-        images = images.cpu().numpy()
+        images = images.cpu().float().numpy()
 
     if isinstance(targets, torch.Tensor):
         targets = targets.cpu().numpy()