segment-anything-tensorrt

segment-anything-tensorrt copied to clipboard

i am using libtorch 2.0.1 , cuda 11.7 , i am running main.cpp with vit_l model the model inference is working fine , getting masks correctly, but failing in verify...

venkat0907

decoder on ONNX for sampleSAM2 project

1

Hi, I also have some difficulties creating the ONNX decoder file for the sampleSAM2 project. I can create the two embedding ONNX files. I see other people struggling with the...

bnascimento

Hi, I got an error at the end of the program. I guess the error related at this line. `eng_0 = std::shared_ptr(new SamEmbedding(std::to_string(1), mEngine, frame));` after I add this: `eng_0.reset();`...

miseonpark0216

TensorRT encountered issues when converting weights between types and that could affect accuracy

3

Hi @mingj2021. When I export the model with `demo.py`, I get the following warning. Is this what you would expect? ``` [TRT] [W] TensorRT encountered issues when converting weights between...

JS901

onnx->trt in C++ get error:this 是 nullptr

5

comment the params "orig_im_size" @torch.no_grad() def forward( self, image_embeddings: torch.Tensor, point_coords: torch.Tensor, point_labels: torch.Tensor, mask_input: torch.Tensor, has_mask_input: torch.Tensor, # orig_im_size: torch.Tensor, ): orig_im_size= torch.tensor([1500, 2250], dtype=torch.float) 在linux中导出onnx模型(TensorRT8.5.1.7) 在windows C++端加载onnx,在进行onnx->trt时报错 nvinfer.h中return...

guanhaiye

How to run updated code

2

Hi @mingj2021, I see that you have updated the repo with new code. Please could you provide usage instructions in the README as it now appears that you have switched...

JS901

转模型报错

2

执行代码: def export_embedding_model(): sam_checkpoint = "D:/Anaconda3/envs/sam_vit_h_4b8939.pth" model_type = "vit_h" device = "cpu" sam = sam_model_registry[model_type](checkpoint=sam_checkpoint) sam.to(device=device) image = cv2.imread('./images/truck.jpg') target_length = sam.image_encoder.img_size pixel_mean = sam.pixel_mean pixel_std = sam.pixel_std img_size =...

kc-w

Hi, Can folks trying this repo comment on the performance improvements they gained on their specific hardware using this approach?

prerakmody

Can't export sam onnx model without "orig_im_size" input

3

Hi mingj2021, I had a problem when using the export_sam_model function in demo.py script: If the "orig_im_size" input in dummy_inputs is commented (as default), I cannot export the onnx model....

jiahan

#tested On Windonws # conda vir envs ``` Package Version -------------------- ------------ absl-py 1.4.0 certifi 2022.12.7 charset-normalizer 2.1.1 colorama 0.4.6 coloredlogs 15.0.1 contourpy 1.0.7 cycler 0.11.0 filelock 3.9.0 flatbuffers 23.3.3...

mingj2021