boxes[:, 0] = torch.where(boxes[:, 0] < threshold, torch.tensor(0,dtype=torch.float,device=boxes.device), boxes[:, 0])  # x1
boxes[:, 1] = torch.where(boxes[:, 1] < threshold, torch.tensor(0,dtype=torch.float,device=boxes.device), boxes[:, 1])  # y1
boxes[:, 2] = torch.where(boxes[:, 2] > w - threshold, torch.tensor(w,dtype=torch.float,device=boxes.device), boxes[:, 2])  # x2
boxes[:, 3] = torch.where(boxes[:, 3] > h - threshold, torch.tensor(h,dtype=torch.float,device=boxes.device), boxes[:, 3])  # y2

is ok

from fastsam.

Hi @syntherick , this appears to be a bug in the code, and we will resolve it as soon as possible. Could you please provide the inference code or command you used? This information would be very helpful for us to investigate the issue.

from fastsam.

Hi @berry-ding, I get the error for either of following:

!python ./Inference.py  --model_path ../FastSAM-x.pt --img_path ./image.jpg --imgsz 1024

and

from fastsam import FastSAM, FastSAMPrompt
model = FastSAM('../FastSAM-x.pt')
IMAGE_PATH = './image.jpg'
DEVICE = '0'
everything_results = model(IMAGE_PATH, device=DEVICE, retina_masks=True, imgsz=1024, conf=0.4, iou=0.9,)

from fastsam.

Hello @syntherick , we have retested and were unable to reproduce the issue. Could you please git pull the latest version of the code and try again? If the problem persists, it might be a bug triggered by specific images. Could you please share the image that caused the problem with us?

from fastsam.

I don't why, but I changed the code in utils.py 17-20, then it works well.

Adjust boxes

boxes[:, 0] = torch.where(boxes[:, 0] < threshold, torch.tensor(0,dtype=torch.float), boxes[:, 0])  # x1
boxes[:, 1] = torch.where(boxes[:, 1] < threshold, torch.tensor(0,dtype=torch.float), boxes[:, 1])  # y1
boxes[:, 2] = torch.where(boxes[:, 2] > w - threshold, torch.tensor(w,dtype=torch.float), boxes[:, 2])  # x2
boxes[:, 3] = torch.where(boxes[:, 3] > h - threshold, torch.tensor(h,dtype=torch.float), boxes[:, 3])  # y2

from fastsam.

@berry-ding Here is the image which I used

from fastsam.

Hi, I have the exactly same error.

from fastsam.

I don't why, but I changed the code in utils.py 17-20, then it works well.

Adjust boxes

boxes[:, 0] = torch.where(boxes[:, 0] < threshold, torch.tensor(0,dtype=torch.float), boxes[:, 0])  # x1
boxes[:, 1] = torch.where(boxes[:, 1] < threshold, torch.tensor(0,dtype=torch.float), boxes[:, 1])  # y1
boxes[:, 2] = torch.where(boxes[:, 2] > w - threshold, torch.tensor(w,dtype=torch.float), boxes[:, 2])  # x2
boxes[:, 3] = torch.where(boxes[:, 3] > h - threshold, torch.tensor(h,dtype=torch.float), boxes[:, 3])  # y2

There will be bugs again again running on cuda devices

from fastsam.

I meet the same proble , when use the inference demo you provide

from fastsam import FastSAM, FastSAMPrompt

model = FastSAM('./weights/FastSAM.pt')
IMAGE_PATH = './images/dogs.jpg'
DEVICE = 'cpu'
everything_results = model(IMAGE_PATH, device=DEVICE, retina_masks=True, imgsz=1024, conf=0.4, iou=0.9,)
prompt_process = FastSAMPrompt(IMAGE_PATH, everything_results, device=DEVICE)

# everything prompt
ann = prompt_process.everything_prompt()

# bbox default shape [0,0,0,0] -> [x1,y1,x2,y2]
ann = prompt_process.box_prompt(bbox=[200, 200, 300, 300])

# text prompt
ann = prompt_process.text_prompt(text='a photo of a dog')

# point prompt
# points default [[0,0]] [[x1,y1],[x2,y2]]
# point_label default [0] [1,0] 0:background, 1:foreground
ann = prompt_process.point_prompt(points=[[620, 360]], pointlabel=[1])

prompt_process.plot(annotations=ann,output='./output/',)

And it show runtimeerror like this
File "inference_demo.py", line 6, in
everything_results = model(IMAGE_PATH, device=DEVICE, retina_masks=True, imgsz=1024, conf=0.4, iou=0.9,)
File "E:\OroChiLab\FastSAM\fastsam\model.py", line 99, in call
return self.predict(source, stream, **kwargs)
File "C:\ProgramData\miniconda3\envs\torch-py37\lib\site-packages\torch\autograd\grad_mode.py", line 27, in decorate_context
return func(*args, **kwargs)
File "E:\OroChiLab\FastSAM\fastsam\model.py", line 51, in predict
return self.predictor(source, stream=stream)
File "C:\ProgramData\miniconda3\envs\torch-py37\lib\site-packages\ultralytics\yolo\engine\predictor.py", line 184, in call
return list(self.stream_inference(source, model)) # merge list of Result into one
File "C:\ProgramData\miniconda3\envs\torch-py37\lib\site-packages\torch\autograd\grad_mode.py", line 43, in generator_context
response = gen.send(None)
File "C:\ProgramData\miniconda3\envs\torch-py37\lib\site-packages\ultralytics\yolo\engine\predictor.py", line 244, in stream_inference
self.results = self.postprocess(preds, im, im0s)
File "E:\OroChiLab\FastSAM\fastsam\predict.py", line 27, in postprocess
critical_iou_index = bbox_iou(full_box[0][:4], p[0][:, :4], iou_thres=0.9, image_shape=img.shape[2:])
File "E:\OroChiLab\FastSAM\fastsam\utils.py", line 50, in bbox_iou
boxes = adjust_bboxes_to_image_border(boxes, image_shape)
File "E:\OroChiLab\FastSAM\fastsam\utils.py", line 19, in adjust_bboxes_to_image_border
boxes[:, 0] = torch.where(boxes[:, 0] < threshold, 0, boxes[:, 0]) # x1
RuntimeError: expected scalar type __int64 but found float

from fastsam.

boxes[:, 0] = torch.where(boxes[:, 0] < threshold, torch.tensor(0,dtype=torch.float,device=boxes.device), boxes[:, 0])  # x1
boxes[:, 1] = torch.where(boxes[:, 1] < threshold, torch.tensor(0,dtype=torch.float,device=boxes.device), boxes[:, 1])  # y1
boxes[:, 2] = torch.where(boxes[:, 2] > w - threshold, torch.tensor(w,dtype=torch.float,device=boxes.device), boxes[:, 2])  # x2
boxes[:, 3] = torch.where(boxes[:, 3] > h - threshold, torch.tensor(h,dtype=torch.float,device=boxes.device), boxes[:, 3])  # y2

is ok

It worked thanks for your answer😋

from fastsam.

We would close this issue for now to keep the issue tracker organized. However, if the problem persists or if you have any further questions, please feel free to comment here or open a new issue. We value your input and are happy to assist further.

from fastsam.

the same problem:

Traceback (most recent call last):
File "Inference.py", line 119, in
main(args)
File "Inference.py", line 79, in main
everything_results = model(
File "/home/rui.li/projects/FastSAM/fastsam/model.py", line 99, in call
return self.predict(source, stream, **kwargs)
File "/home/rui.li/projects/FastSAM/venv/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context
return func(*args, **kwargs)
File "/home/rui.li/projects/FastSAM/fastsam/model.py", line 51, in predict
return self.predictor(source, stream=stream)
File "/home/rui.li/projects/FastSAM/venv/lib/python3.8/site-packages/ultralytics/yolo/engine/predictor.py", line 188, in call
return list(self.stream_inference(source, model)) # merge list of Result into one
File "/home/rui.li/projects/FastSAM/venv/lib/python3.8/site-packages/torch/autograd/grad_mode.py", line 45, in generator_context
response = gen.send(None)
File "/home/rui.li/projects/FastSAM/venv/lib/python3.8/site-packages/ultralytics/yolo/engine/predictor.py", line 248, in stream_inference
self.results = self.postprocess(preds, im, im0s)
File "/home/rui.li/projects/FastSAM/fastsam/predict.py", line 31, in postprocess
critical_iou_index = bbox_iou(full_box[0][:4], p[0][:, :4], iou_thres=0.9, image_shape=img.shape[2:])
File "/home/rui.li/projects/FastSAM/fastsam/utils.py", line 39, in bbox_iou
boxes = adjust_bboxes_to_image_border(boxes, image_shape)
File "/home/rui.li/projects/FastSAM/fastsam/utils.py", line 17, in adjust_bboxes_to_image_border
boxes[:, 0] = torch.where(boxes[:, 0] < threshold, 0, boxes[:, 0]) # x1
RuntimeError: expected scalar type long int but found float

from fastsam.

我不知道为什么，但我更改了 utils.py 17-20 中的代码，然后它就可以正常工作了。

调整盒

boxes[:, 0] = torch.where(boxes[:, 0] < threshold, torch.tensor(0,dtype=torch.float), boxes[:, 0])  # x1
boxes[:, 1] = torch.where(boxes[:, 1] < threshold, torch.tensor(0,dtype=torch.float), boxes[:, 1])  # y1
boxes[:, 2] = torch.where(boxes[:, 2] > w - threshold, torch.tensor(w,dtype=torch.float), boxes[:, 2])  # x2
boxes[:, 3] = torch.where(boxes[:, 3] > h - threshold, torch.tensor(h,dtype=torch.float), boxes[:, 3])  # y2

Thank you! we have fixed our codes.

from fastsam.