run detect.py output nothing

[jedi008]

OS: windows10 GPU: NVIDIA GeForce MX450 Command: python detect.py Weights: https://github.com/WongKinYiu/yolov7/releases/download/v0.1/yolov7.pt and yolov7x.pt Output: didn't detect any object.

The reason is that pred is an invalid value: https://github.com/WongKinYiu/yolov7/blob/55b90e111984dd85e7eed327e9ff271222aa8b82/detect.py#L87 pred is [nan, nan, nan, ..., nan, nan, nan].

Change this line to： half = False https://github.com/WongKinYiu/yolov7/blob/55b90e111984dd85e7eed327e9ff271222aa8b82/detect.py#L31 Then got the right result.

[whittenator]

I am receiving the same issue. When running detect.py it isn't saving anything. I run the following command: python detect.py --device 0 --weights runs/train/yolov7-xod335/weights/best.pt --conf 0.25 --img-size 416 --source inference/videos/MoogIR-15-11-52_333.mp4 --name xod335-test

And then I receive the following output:

Namespace(agnostic_nms=False, augment=False, classes=None, conf_thres=0.25, device='0', exist_ok=False, img_size=416, iou_thres=0.45, name='xod335-test', no_trace=False, nosave=False, project='runs/detect', save_conf=False, save_txt=False, source='inference/videos/MoogIR-15-11-52_333.mp4', update=False, view_img=False, weights=['runs/train/yolov7-xod335/weights/best.pt'])
YOLOR 🚀 2022-9-20 torch 1.13.0a0+d321be6 CUDA:0 (NVIDIA GeForce RTX 3080, 10010.0MB)

Fusing layers... 
RepConv.fuse_repvgg_block
RepConv.fuse_repvgg_block
RepConv.fuse_repvgg_block
IDetect.fuse
/opt/conda/lib/python3.8/site-packages/torch/functional.py:478: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/pytorch/pytorch/aten/src/ATen/native/TensorShape.cpp:2966.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
Model Summary: 314 layers, 36487166 parameters, 6194944 gradients, 103.2 GFLOPS
 Convert model to Traced-model... 
 traced_script_module saved! 
 model is traced! 

Done. (0.000s)

However, my runs/detect/xod335-test folder is completely empty. Any suggestions?

[mkrupczak3]

Same issue

[whittenator]

I found out the issue, the PyTorch Docker image that they are using is using an OpenCV that doesn't have Video I/O installed. I found that installing OpenCV with these cleared my issue!

[mkrupczak3]

@whittenator awesome, could you elaborate what steps you ran to get the container working for inference on video?

[Raphael138]

@whittenator Yo big guy I’m also with mkrupczak3 and would love a quick tutorial or set of direction.

[Garbage123King]

I use docker, and this helps me. Thanks whittenator.

pip uninstall opencv-python
pip uninstall opencv-contrib-python
pip install opencv_python==3.4.10.37

[whittenator]

@whittenator Yo big guy I’m also with mkrupczak3 and would love a quick tutorial or set of direction.

Sorry for the delay fellas, I had to create a docker image with a custom installation of OpenCV. If you are like me and would like to run inference on your Nvidia GPU then you can use the Docker I created below. Note: Make sure you select the correct arch type for your Nvidia GPU Here is the Docker:

FROM nvcr.io/nvidia/pytorch:22.08-py3

ARG DEBIAN_FRONTEND=noninteractive
ARG OPENCV_VERSION=4.6.0

RUN apt-get update && apt-get upgrade -y &&\
    # Install build tools, build dependencies and python
    apt-get install -y \
    python3-pip \
        build-essential \
        cmake \
        git \
        wget \
        unzip \
        yasm \
        pkg-config \
        libswscale-dev \
        libgtk-3-dev \
        libx264-dev \
        libtbb2 \
        libtbb-dev \
        libjpeg-dev \
        libpng-dev \
        libtiff-dev \
        libv4l-dev \
        gfortran \
        openexr \
        libdc1394-22-dev \
        libgstreamer-plugins-base1.0-dev \
        libgstreamer1.0-dev \
        libopenexr-dev \
        libatlas-base-dev \
        libxvidcore-dev \
        libavformat-dev \
        libpq-dev \
        libxine2-dev \
        libglew-dev \
        libtiff5-dev \
        zlib1g-dev \
        libavcodec-dev \
        libavutil-dev \
        libpostproc-dev \
        libswscale-dev \
        libeigen3-dev \
        libtbb-dev \
        libgtk2.0-dev \
        ffmpeg \
        pkg-config \
        ## Python
        python3-dev \
        python3-numpy \
        python3-magic \
        zip \
        htop \
        screen \ 
        libgl1-mesa-glx \
    && rm -rf /var/lib/apt/lists/*

# Install all Video I/O and OpenCV Dependencies
RUN apt-get install -y \
    build-essential cmake git pkg-config libgtk-3-dev \
    libavcodec-dev libavformat-dev libswscale-dev libv4l-dev \
    libxvidcore-dev libx264-dev libjpeg-dev libpng-dev libtiff-dev \
    gfortran openexr libatlas-base-dev python3-dev python3-numpy \
    libtbb2 libtbb-dev libdc1394-22-dev libopenexr-dev \
    libgstreamer-plugins-base1.0-dev libgstreamer1.0-dev

RUN cd /opt/ &&\
    # Download and unzip OpenCV and opencv_contrib and delte zip files
    wget https://github.com/opencv/opencv/archive/$OPENCV_VERSION.zip &&\
    unzip $OPENCV_VERSION.zip &&\
    rm $OPENCV_VERSION.zip &&\
    wget https://github.com/opencv/opencv_contrib/archive/$OPENCV_VERSION.zip &&\
    unzip ${OPENCV_VERSION}.zip &&\
    rm ${OPENCV_VERSION}.zip &&\
    # Create build folder and switch to it
    mkdir /opt/opencv-${OPENCV_VERSION}/build && cd /opt/opencv-${OPENCV_VERSION}/build &&\
    # Cmake configure
    cmake \
        -DOPENCV_EXTRA_MODULES_PATH=/opt/opencv_contrib-${OPENCV_VERSION}/modules \
        -DWITH_CUDA=ON \
        -DCUDA_GENERATION=Auto \
    -DWITH_CUDNN=ON \
    -DWITH_CUBLAS=ON \
    -DWITH_TBB=ON \
    -DOPENCV_DNN_CUDA=ON \
    -DOPENCV_ENABLE_NONFREE=ON \
        -DCUDA_ARCH_BIN=7.5,8.0,8.6 \
        -DWITH_FFMPEG=ON
        -DWITH_GSTREAMER=ON
        -D HAVE_opencv_python3=ON \
        -DCMAKE_BUILD_TYPE=RELEASE \
        # Install path will be /usr/local/lib (lib is implicit)
        -DCMAKE_INSTALL_PREFIX=/usr/local \
        .. &&\
    # Make
    make -j"$(nproc)" && \
    # Install to /usr/local/lib
    make install && \
    ldconfig &&\
    # Remove OpenCV sources and build folder
    rm -rf /opt/opencv-${OPENCV_VERSION} && rm -rf /opt/opencv_contrib-${OPENCV_VERSION}

# Install python dependencies    
RUN pip install seaborn thop nvidia-pyindex nvidia-tensorrt pycuda

# The YoloV7 training repo
RUN git clone https://github.com/WongKinYiu/yolov7.git

# Clone repo that allows conversion of pytorch to TensorRT with NMS
RUN git clone https://github.com/Linaom1214/tensorrt-python.git 

Hope this helps!