def evaluate(data_loader):
meter = Meter('eval', 0)
model.eval()
total_loss = 0
with torch.no_grad():
for idx, (img, segm) in enumerate(data_loader):
img = img.cuda()
segm = segm.cuda()
outputs = model(img)
loss = criterion(outputs, segm)
outputs = outputs.detach().cpu()
segm = segm.detach().cpu()
meter.update(segm, outputs)
total_loss += loss.item()
dices, iou = meter.get_metrics()
dice, dice_neg, dice_pos = dices
torch.cuda.empty_cache()
return total_loss/len(data_loader), iou, dice, dice_neg, dice_pos
alpha = 0.5
if (TARGET == "kitti") and (orig_image.shape[:2] != pic.shape[:2]):
orig_image = cv2.resize(
orig_image,
(DATASET["orig_size"][1], DATASET["orig_size"][0]),
cv2.INTER_LANCZOS4)
pic = cv2.addWeighted(orig_image, (1 - alpha), pic, alpha, 0)
pred_name = "_".join(
re.split("\.|_",
image_id[0].split("/")[-1])[:-1]) + "_predicted_mask.png"
cv2.imwrite(os.path.join(images_path, pred_name), pic)
torch.cuda.empty_cache()
if not EVAL["test_mode"]:
dices, iou = meter.get_metrics("val")
print("***** Prediction done in {} sec.; IoU: {}, Dice: {} ***** \n(total elapsed time: {} sec.) ".\
format(int(time.time()-start), iou, dices[0]["dice_all"], int(time.time()-global_start)))
if TARGET == "cityscapes" and len(dices[0]) > 1:
labels_df = image_dataset.label_encoder.cityscapes_labels_df
if DATASET["train_on_cats"]:
cat, name = "catId", "category"
else:
cat, name = "trainId", "name"
print("***** Class metrics: *****")
for k, v in dices[0].items():
if k != "dice_all":
print(labels_df[labels_df[cat] == int(k)][name].iloc[0],
" : ", v)
else:
print("***** Prediction on test set done in {} sec. ***** \n(total elapsed time: {} sec.) ".\