def export(self):
shape = (1, 3, self.size[0], self.size[1])
yolo_gluon.export(self.net,
shape,
self.ctx[0],
self.export_file,
onnx=1,
epoch=0)
def export(self):
shape = (1, 3, self.size[0], self.size[1])
yolo_gluon.export(self.net,
shape,
self.ctx[0],
self.export_folder,
onnx=False,
fp16=self.use_fp16)
def export(self):
yolo_gluon.export(
self.net,
(1, 3, self.size[0], self.size[1]),
self.ctx[0],
self.export_folder,
onnx=False,
fp16=False)
bg = bg.data[0].as_in_context(ctx[0])
imgs, labels = generator.render(bg)
score_x, class_x = net(imgs)
print(score_x.shape)
print(class_x.shape)
imgs = yolo_gluon.batch_ndimg_2_cv2img(imgs)
for i in range(bs):
ax = axs[i]
s = score_x[i]
s = nd.sigmoid(s.reshape(-1)).asnumpy()
p = class_x[i, 0].asnumpy()
p = np.argmax(p, axis=-1)
yolo_cv.matplotlib_show_img(ax, imgs[i])
ax.plot(range(8, 384, 16), (1 - s) * 160)
ax.axis('off')
s = np.concatenate(([0], s, [0]))
# zero-dimensional arrays cannot be concatenated
# Find peaks
text = ''
for i in range(24):
if s[i + 1] > 0.2 and s[i + 1] > s[i + 2] and s[i + 1] > s[i]:
c = int(p[i])
text = text + cls_names[c]
print(text)
raw_input('press Enter to next batch....')
elif args.mode == 'export':
yolo_gluon.export(net, (1, 3, size[0], size[1]), ctx[0], export_file)
def export(self):
yolo_gluon.export(self.net, (1, 3, self.size[0], self.size[1]),
self.ctx[0],
self.export_file,
onnx=0,
epoch=0)