Using the provided model file to test the picture has an error.

[assertdebug]

[assertdebug]

Sorry , I forgot to specify the input size

[assertdebug]

I use the model file provided and cannot get the correct result

[assertdebug]

I use the model file provided and cannot get the correct result

[gjhhust]

I use the model file provided and cannot get the correct result

Currently does not support the direct use of yolo commands to reason, please use a script to get the results, and YOLOFT model is suitable for video reasoning, You need to use the build_stream_dataloader input model for video detection. You can refer to https://github.com/gjhhust/YOLOFT/blob/main/tools/test_flowft_show.py

[Yu-zhengbo]

I use the model file provided and cannot get the correct result

Currently does not support the direct use of yolo commands to reason, please use a script to get the results, and YOLOFT model is suitable for video reasoning, You need to use the build_stream_dataloader input model for video detection. You can refer to https://github.com/gjhhust/YOLOFT/blob/main/tools/test_flowft_show.py

你好，看了你写的代码，build_stream_dataloader好像返回的也是一帧一帧的图像，因为看了你的论文里面描述训练阶段是多帧输入，所以想请教一下是我理解错了吗

[gjhhust]

I wrote predict scripts for images and videos, For more information, please read readme.md.

and Sorry only just noticed your question, the paper describes streaming input, i.e. only one image input at a point in time (consistent with real world acquisition of images)