zhudongwork

Yes, you can refer to https://github.com/dinglufe/segment-anything-cpp-wrapper.

Yes, you can refer to https://github.com/dinglufe/segment-anything-cpp-wrapper.

Yes, the input size of the model is 1024*1024. The images were processed using ResizeLongestSide-1024 in the original SAM project. In order to avoid tensorrt's dynamic inference of orig_img_size, the...

Please try TensorRT-8.5

This is confusing, I tried it on both TensorRT-8.6.1.6 and 8.5.3.1 and it works fine. This may be an environmental problem or hardware issue with your local tensorrt.

In main.cpp, the upper left corner of cv::Rect is used as the prompt point of sam. In your case1, the upper left corner belongs to the ground, which leads to...

> 微调的数据集不含response字段嘛 有的，这是数据的样式： 

微调Cogvlm2的时候也是一样的问题，是我数据集的格式不对吗。 这是运行命令 `CUDA_VISIBLE_DEVICES=7 swift sft --model_id_or_path /node6/models/ZhipuAI/cogvlm2-llama3-chinese-chat-19B --model_type cogvlm2-19b-chat --dataset ../ocr_100.json --batch_size 4 --val_dataset_sample 10` 微调数据 [ocr_100.json](https://github.com/user-attachments/files/15510763/ocr_100.json)

> > > Please download the pdf parser related checkpoints in modelscope [https://www.modelscope.cn/models/netease-youdao/QAnything-pdf-parser/files] > > > > > > 好的十分感谢，另外是不是Qanything无法处理没有文本元素的pdf啊，我截了一张图进行解析，发现有报错。如果是这样那它里面的ocr的意义是什么呢，是解析表格？ > > 报错信息如下: LOCAL DATA PATH: c:\Users\Administrator\Desktop\QAnything-1.4.1\QANY_DB\content LOCAL_RERANK_REPO: netease-youdao/bce-reranker-base_v1 LOCAL_EMBED_REPO:...

> from qanything_kernel.utils.loader.self_pdf_loader import PdfLoader pdf_loader = PdfLoader(filename='tables/table-03d9ec345317b0115180d7dbcf843ef6.pdf') markdown_directory = pdf_loader.load_to_markdown() print(f"Markdown文件在: {markdown_directory}") > > ➜ QAnything python QAnything_ocr.py LOCAL DATA PATH: /mnt/user/QAnything-qanything-python/QAnything/QANY_DB/content LOCAL_RERANK_REPO: netease-youdao/bce-reranker-base_v1 LOCAL_EMBED_REPO: netease-youdao/bce-embedding-base_v1 Traceback (most recent...