def display_image_depth(image, depth):
depth.add_(-depth.min())
depth = depth.div_(1000 / 256).clamp(max=255).byte()
depth = cv.gray_to_rgb(depth)
cv.display(torch.cat([image, depth], 1))
classes = model_args.dataset.classes
model.to(device)
encoder.to(device)
frame = cv.imread_color(args.input)
nms_params = detection_table.nms_defaults._extend(nms=args.threshold)
pprint_struct(nms_params)
detections = evaluate_image(model,
frame,
encoder,
nms_params=nms_params,
device=device,
crop_boxes=True)
for prediction in detections._sequence():
if prediction.confidence > 0.7:
label_class = classes[prediction.label].name
display.draw_box(frame,
prediction.bbox,
confidence=prediction.confidence,
name=label_class.name,
color=display.to_rgb(label_class.colour))
frame = cv.resize(frame, (frame.size(1) // 2, frame.size(0) // 2))
cv.display(frame)
def show_indexed_batch(t, cols=int(6)):
colorizer = index_map.colorizer_t(255)
tiled = tile_batch(t, cols)
color = colorizer(tiled)
return cv.display(color)
def show_batch(t, cols=int(6), scale=1):
tiled = tile_batch(t, cols)
tiled = cv.resize (tiled, (tiled.size(0) * scale, tiled.size(1) * scale), interpolation = cv.INTER_NEAREST)
return cv.display(tiled)
transform=random_crop(dim))
loader=DataLoader(dataset, num_workers=args.num_workers,
batch_size=args.batch_size, sampler=RepeatSampler(args.epoch_size, dataset))
return loader, dataset
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Segmentation - view data set')
parser.add_argument('--input', default='/storage/workspace/dtd/images/scaly',
help='input image path')
args = parser.parse_args()
args.num_workers = 1
args.epoch_size = 1024
args.batch_size = 16
loader, dataset = training(args, (144, 144))
for _, data in enumerate(loader):
image = tensor.tile_batch(data, cols=4)
if (cv.display(image) == 27):
break
def visualise(model, encoder, iter, args):
threshold = 50
zoom = 100
modes = ['target', 'matches', 'prediction', 'normal']
mode_index = 0
debug_index = 0
help_str = """
keys:
-+: reduce and increase threshold
[]: zoom in and out
(): adjust nms threshold
<>: adjust max detections
m: show matches
p: show predictions
a: show matched anchor boxes
t: show target boxes
"""
keys = struct(escape=27, space=32, minus=45, plus=61)
nms = struct(nms=0.5, threshold=0.05, detections=400)
print(help_str)
model.to(device)
encoder.to(device)
print("classes:")
print(dataset.classes)
for batch in iter:
key = 0
while (not key in [keys.escape, keys.space]):
debug_key = encoder.debug_keys[debug_index -
1] if debug_index > 0 else None
evals = [
evaluate_vis(model,
encoder,
i,
nms,
dataset.classes,
args,
debug_key=debug_key) for i in batch
]
def show2(x):
return "{:.2f}".format(x)
def show_vector(v):
return "({})".format(",".join(map(show2, v.tolist())))
for e in evals:
print("{}: {:d}x{:d} {}".format(
e.file, e.image_size[0], e.image_size[1],
str(e.stats._map(show_vector))))
display = overlay_batch(evals,
mode=modes[mode_index],
scale=100 / zoom,
classes=dataset.classes,
threshold=threshold / 100,
cols=4)
if zoom != 100:
display = resize_scale(display, zoom / 100)
key = cv.display(display)
if key == keys.minus and threshold > 0:
threshold -= 5
print("reducing threshold: ", threshold)
elif key == keys.plus and threshold < 100:
threshold += 5
print("increasing threshold: ", threshold)
elif chr(key) == '(' and nms.nms > 0:
nms.nms = max(0, nms.nms - 0.05)
print("decreasing nms threshold: ", nms.nms)
elif chr(key) == ')' and nms.nms < 1:
nms.nms = min(1, 0.05 + nms.nms)
print("increasing nms threshold: ", nms.nms)
elif chr(key) == '<' and nms.detections > 50:
nms.detections -= 50
print("decreasing max detections: ", nms.detections)
elif chr(key) == '>':
nms.detections += 50
print("increasing max detections: ", nms.detections)
elif key == 91 and zoom > 10:
zoom -= 5
print("zoom out: ", zoom)
elif key == 93 and zoom < 200:
zoom += 5
print("zoom in: ", zoom)
elif chr(key) == 'm':
mode_index = (mode_index + 1) % len(modes)
print("showing " + modes[mode_index])
elif chr(key) == 'd':
debug_index = (debug_index + 1) % (len(encoder.debug_keys) + 1)
if debug_index > 0:
print("showing debug: " +
encoder.debug_keys[debug_index - 1])
else:
print("hiding debug")
if (key == 27):
break