def inference(image_path, keep_thresh=0.5, savepath=None):
"""
Args:
image_path: image path
keep_thresh: 置信度大于多少的保留
savepath: 预览图保存路径，为None不保存
Returns:
a tuple (bboxes, labels, scores)
bboxes:
[[x1, y1, x2, y2], [x1, y1, x2, y2], ...]
labels:
[label1, label2, ...]
scores:
[score1, score2, ...] # 降序排列
"""
assert os.path.isfile(image_path), f'{image_path} not exists.'
filaname = utils.get_file_name(image_path)
image_org = cv2.imread(image_path)
image_org = cv2.cvtColor(image_org, cv2.COLOR_BGR2RGB).astype(np.float32)
image = image_org / 255.0
sample = test_transform(**{
'image': image,
})
image = sample['image']
image = image.unsqueeze(0).to(opt.device)
batch_bboxes, batch_labels, batch_scores = model.forward_test(image)
bboxes = batch_bboxes[0]
labels = batch_labels[0]
scores = batch_scores[0]
keep = scores > keep_thresh
bboxes = bboxes[keep]
labels = labels[keep]
scores = scores[keep]
R = lambda x: int(round(x, 0)) # 四舍五入
visualize_boxes(image=image_org,
boxes=bboxes,
labels=labels,
probs=scores,
class_labels=class_names)
bboxes_arr = []
for i in range(len(scores)):
x1, y1, x2, y2 = bboxes[i]
x1, y1, x2, y2 = R(x1), R(y1), R(x2), R(y2)
bboxes_arr.append([x1, y1, x2, y2])
bboxes = np.array(bboxes_arr)
if savepath:
print(f'save image to {savepath}/{filaname}.png')
image = cv2.cvtColor(np.asarray(image_org), cv2.COLOR_RGB2BGR)
cv2.imwrite(f'{savepath}/{filaname}.png', image)
return bboxes, labels, scores
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB).astype(np.float32)
img /= 255.0 # 转成0~1之间
img_input = val_transform(image=img)['image']
img_input = img_input.unsqueeze(0).to(opt.device)
batch_bboxes, batch_labels, batch_scores = model(img_input)
img = tensor2im(img_input).copy()
# for x1, y1, x2, y2 in gt_bbox[0]:
# cv2.rectangle(img, (x1,y1), (x2,y2), (0, 255, 0), 2) # 绿色的是gt
num = len(batch_scores[0])
visualize_boxes(image=img,
boxes=batch_bboxes[0],
labels=batch_labels[0].astype(np.int32),
probs=batch_scores[0],
class_labels=config.DATA.CLASS_NAMES)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# img=np.rot90(img)
frames = frames + 1
nt = time.time()
if nt - t > 1.0:
fps = frames
frames = 0
t = nt
# Draw fps
def eval_mAP(self, dataloader, epoch, writer, logger, data_name='val'):
# eval_yolo(self.detector, dataloader, epoch, writer, logger, dataname=data_name)
pred_bboxes = []
pred_labels = []
pred_scores = []
gt_bboxes = []
gt_labels = []
gt_difficults = []
with torch.no_grad():
for i, sample in enumerate(dataloader):
utils.progress_bar(i, len(dataloader), 'Eva... ')
image = sample['image'].to(opt.device)
gt_bbox = sample['bboxes']
labels = sample['labels']
paths = sample['path']
batch_bboxes, batch_labels, batch_scores = self.forward(image)
pred_bboxes.extend(batch_bboxes)
pred_labels.extend(batch_labels)
pred_scores.extend(batch_scores)
for b in range(len(gt_bbox)):
gt_bboxes.append(gt_bbox[b].detach().cpu().numpy())
gt_labels.append(labels[b].int().detach().cpu().numpy())
gt_difficults.append(np.array([False] * len(gt_bbox[b])))
if opt.vis: # 可视化预测结果
img = tensor2im(image).copy()
# for x1, y1, x2, y2 in gt_bbox[0]:
# cv2.rectangle(img, (x1,y1), (x2,y2), (0, 255, 0), 2) # 绿色的是gt
num = len(batch_scores[0])
visualize_boxes(image=img,
boxes=batch_bboxes[0],
labels=batch_labels[0].astype(np.int32),
probs=batch_scores[0],
class_labels=opt.class_names)
write_image(writer, f'{data_name}/{i}', 'image', img,
epoch, 'HWC')
result = []
for iou_thresh in [0.5, 0.55, 0.6, 0.65, 0.7, 0.75]:
AP = eval_detection_voc(pred_bboxes,
pred_labels,
pred_scores,
gt_bboxes,
gt_labels,
gt_difficults=None,
iou_thresh=iou_thresh,
use_07_metric=False)
APs = AP['ap']
mAP = AP['map']
result.append(mAP)
logger.info(
f'Eva({data_name}) epoch {epoch}, IoU: {iou_thresh}, APs: {str(APs[:opt.num_classes])}, mAP: {mAP}'
)
write_loss(writer, f'val/{data_name}', 'mAP', mAP, epoch)
logger.info(
f'Eva({data_name}) epoch {epoch}, mean of (AP50-AP75): {sum(result)/len(result)}'
)
# class_names = opt.class_names
class_names = opt.class_names
preview = val_dataloader # train_dataloader, val_dataloader
from utils.vis import visualize_boxes
for i, sample in enumerate(preview):
# if i > 30:
# break
utils.progress_bar(i, len(preview), 'Handling...')
if opt.debug:
ipdb.set_trace()
image = sample['image'][0].detach().cpu().numpy().transpose([1, 2, 0])
image = (image.copy() * 255).astype(np.uint8)
bboxes = sample['bboxes'][0].cpu().numpy()
labels = sample['labels'][0].cpu().numpy().astype(np.int32)
visualize_boxes(image=image,
boxes=bboxes,
labels=labels,
probs=np.array(
np.random.randint(100, 101, size=[len(bboxes)]) / 100),
class_labels=class_names)
write_image(writer, f'preview_{opt.dataset}/{i}', 'image', image, 0, 'HWC')
writer.flush()
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB).astype(np.float32)
img /= 255.0 # 转成0~1之间
img_input = val_transform(image=img)['image']
img_input = img_input.unsqueeze(0).to(opt.device)
batch_bboxes, batch_labels, batch_scores = model(img_input)
img = tensor2im(img_input).copy()
# for x1, y1, x2, y2 in gt_bbox[0]:
# cv2.rectangle(img, (x1,y1), (x2,y2), (0, 255, 0), 2) # 绿色的是gt
num = len(batch_scores[0])
visualize_boxes(image=img,
boxes=batch_bboxes[0],
labels=batch_labels[0].astype(np.int32),
probs=batch_scores[0],
class_labels=opt.class_names)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# img=np.rot90(img)
frames = frames + 1
nt = time.time()
if nt - t > 1.0:
fps = frames
frames = 0
t = nt
# Draw fps