kenho211

Bounding box in the same line break down into smaller ones after finetuning

4

Hi everyone, I have used a custom dataset (forms and documents) to finetune on chinese+english detection, using the following: config: ch_PP-OCRv3_det_student.yml pretrain_model: ./pretrain_models/ch_PP-OCRv3_det_distill_train/student Using pretrained model, the detected text in...

Wondering if the documentation of samples_width is correct here. https://github.com/alesaccoia/VoiceStreamAI/blob/465403b7039d1f54ba6b8d69c69c40b55bf300c1/src/server.py#L25 Because when calculating for buffer and scratch_buffer, there is no division by 8. https://github.com/alesaccoia/VoiceStreamAI/blob/465403b7039d1f54ba6b8d69c69c40b55bf300c1/src/buffering_strategy/buffering_strategies.py#L116-L118 https://github.com/alesaccoia/VoiceStreamAI/blob/465403b7039d1f54ba6b8d69c69c40b55bf300c1/src/buffering_strategy/buffering_strategies.py#L73-L77

https://github.com/IDEA-Research/GroundingDINO/blob/856dde20aee659246248e20734ef9ba5214f5e44/demo/inference_on_a_image.py#L120C29-L120C40 in `demo/inference_on_a_image.py`, `text_prompt` is used `def get_grounding_output` but it is not defined in the function. Should use `caption` instead

Incompatible transformers version in requirements.txt

2

ImportError: cannot import name 'apply_chunking_to_forward' from 'transformers.modeling_utils' (/opt/venv/image-recognition/lib/python3.11/site-packages/transformers/modeling_utils.py) In requirements.txt, transformers>=4.25.1 is used. But after transformers v4.55.4, `apply_chunking_to_forward` is not imported from pytorch_utils anymore.

To solve https://github.com/xinyu1205/recognize-anything/issues/219