How to enter multiple coordinates?

[weijiafs]

Hi, I have a question, how to enter multiple coordinates?

python remove_anything.py \

--input_img ~/input/test1.png \
--point_coords 58 127 778 127 58 366 778 366 \
--point_labels 4 \
--dilate_kernel_size 15 \
--output_dir ~/output \
--sam_model_type "vit_h" \
--sam_ckpt ~/sam/sam_vit_h_4b8939.pth \
--lama_config ./lama/configs/prediction/default.yaml \
--lama_ckpt big-lama

Traceback (most recent call last): File "/root/Inpaint-Anything/remove_anything.py", line 75, in masks, _, _ = predict_masks_with_sam( File "/root/Inpaint-Anything/sam_segment.py", line 28, in predict_masks_with_sam masks, scores, logits = predictor.predict( File "/root/Inpaint-Anything/segment_anything/segment_anything/predictor.py", line 154, in predict masks, iou_predictions, low_res_masks = self.predict_torch( File "/root/miniconda3/lib/python3.9/site-packages/torch/autograd/grad_mode.py", line 28, in decorate_context return func(*args, **kwargs) File "/root/Inpaint-Anything/segment_anything/segment_anything/predictor.py", line 222, in predict_torch sparse_embeddings, dense_embeddings = self.model.prompt_encoder( File "/root/miniconda3/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1102, in _call_impl return forward_call(*input, **kwargs) File "/root/Inpaint-Anything/segment_anything/segment_anything/modeling/prompt_encoder.py", line 155, in forward point_embeddings = self._embed_points(coords, labels, pad=(boxes is None)) File "/root/Inpaint-Anything/segment_anything/segment_anything/modeling/prompt_encoder.py", line 84, in _embed_points points = torch.cat([points, padding_point], dim=1) RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 8 but got size 2 for tensor number 1 in the list.

[RuoyuFeng]

Hello, may I ask about the reason for wanting to input multiple coordinates? Are you attempting to refine the mask of a single object?" And currently, we only implement the function with one coordinate as input~

[weijiafs]

Hello

I want to mark more accurately by multiple coordinate points. This display supports multiple coordinate points. Will it support multi-coordinate points in the future?

usage: remove_anything.py [-h] --input_img INPUT_IMG --point_coords POINT_COORDS [POINT_COORDS ...] --point_labels POINT_LABELS [POINT_LABELS ...] [--dilate_kernel_size DILATE_KERNEL_SIZE] --output_dir OUTPUT_DIR [--sam_model_type {vit_h,vit_l,vit_b}] --sam_ckpt SAM_CKPT [--lama_config LAMA_CONFIG] --lama_ckpt LAMA_CKPT

[RuoyuFeng]

Hello

I want to mark more accurately by multiple coordinate points. This display supports multiple coordinate points. Will it support multi-coordinate points in the future?

usage: remove_anything.py [-h] --input_img INPUT_IMG --point_coords POINT_COORDS [POINT_COORDS ...] --point_labels POINT_LABELS [POINT_LABELS ...] [--dilate_kernel_size DILATE_KERNEL_SIZE] --output_dir OUTPUT_DIR [--sam_model_type {vit_h,vit_l,vit_b}] --sam_ckpt SAM_CKPT [--lama_config LAMA_CONFIG] --lama_ckpt LAMA_CKPT

Hello! Thank you so much for your suggestion. We're happy and willing to add support for multi-coordinate points soon, which will greatly improve the accuracy of our mask generation. Stay tuned for updates!

[chenjun0210]

looking forward to support for multi-coordinate points soon

[geekyutao]

Thank you. Will be supported.

[ChamnLee]

Hi, Does the multi-coordinate points come? Is it feasible for me to try to record multiple coordinate points using a csv file and then iterate to read the points for processing？

[tsqt2023]

2024年2月120.期待官方的多座标点的更新！！