using compute_metrics.py for segment anything

[XiangyuLijoey]

I want to evaluate the results of segment anything on cells. The issue is that I am struggling to convert the masks generated by segment anything to the format required by compute_metrics.py. I followed the examples:

def generate(self, image: np.ndarray) -> List[Dict[str, Any]]:
        """
        Generates masks for the given image.

        Arguments:
          image (np.ndarray): The image to generate masks for, in HWC uint8 format.

        Returns:
           list(dict(str, any)): A list over records for masks. Each record is
             a dict containing the following keys:
               segmentation (dict(str, any) or np.ndarray): The mask. If
                 output_mode='binary_mask', is an array of shape HW. Otherwise,
                 is a dictionary containing the RLE.`

This is the definition of "segmentation" in the generate function of segment anything.

for name in names:
    # 1. read image: img_data = imread(os.path.join(input_path, name))
    # 2. infer the image with your model (the output should be instance mask) 
    inst_mask = model(img_data)
    # relabel the mask
    inst_mask, _, _ = segmentaiton.relabel_sequential(inst_mask)
    # save results; Please compress the mask by setting the `compression`
    save_name = name.split('.')[0] + '_label.tiff' # the suffix of the segmentation results should be '_label.tiff'
    tifffile.imwrite(join(output_path, save_name), np.int16(inst_mask), compression='zlib') 

I checked the example code for how to handle the data in the dataset but still failed to make it work.

[JunMa11]

Hi @XiangyuLijoey ,

Here is a demo to save the SAM auto-seg results to the instance label map.

    masks = mask_generator.generate(image)
    mask_np = np.zeros((image.shape[0], image.shape[1]), dtype=np.uint16)
    for idx, mask in enumerate(masks, start=1):
        mask_np[mask['segmentation']] = idx 
    mask_np, _, _ = segmentation.relabel_sequential(mask_np)
    # remove small regions
    tif.imwrite(join(save_path, name.replace('.png', '.tiff')), mask_np.astype(np.uint16))