MS-DAYOLO

This repository provides a compiled version of MS-DAYOLO for Windows 10, based on the original project:
GitHub - Mazin-Hnewa/MS-DAYOLO

---

Prerequisites

Here's something must install, please follow Preparation Steps:

• Visual Studio 2017 or 2019 (for CUDA Toolkit 10.1 installation)
• Windows 10
• OpenCV 4.9
• CUDA Toolkit 10.1
• cuDNN 7.6.5 (compatible with CUDA 10.1)

---

Preparation Steps

1. Install Visual Studio

Download and install Visual Studio 2017 or 2019 with the Desktop Development with C++ workload.
👉 Visual Studio 2019 Release Notes

---

2. Install CUDA Toolkit 10.1

Download and install CUDA Toolkit 10.1 from the CUDA Toolkit Archive.

Note: Make sure Visual Studio is installed first.

---

3. Set Environment Variables

Add the following system environment variables:

System Variable	Value
CUDA_PATH	C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1
CUDA_PATH_V10_1	C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1

Update the Path environment variable by adding these entries:

Path
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\libnvvp

---

4. Verify CUDA Installation

Open a Command Prompt and run the following command to verify the installation:

nvcc --version

You should see output similar to this:

---

5. Install cuDNN 7.6.5

Download cuDNN 7.6.5 for CUDA 10.1 from the cuDNN Archive.
After extraction, you should see files like this:

Copy the files to their respective directories in your CUDA installation path (C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1):

cuDNN File	Target Directory
bin\cudnn64_7.dll	C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin
include\cudnn.h	C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include
lib\x64\cudnn.lib	C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\lib\x64

---

6. Install OpenCV 4.9

Download OpenCV 4.9 from the OpenCV Releases page.
Run the .exe and extract the files to C:\opencv.

Add the following system environment variable:

System Variable	Value
OpenCV_DIR	C:\opencv\build

Copy the following files to build\darknet\x64:

File	Source Directory
opencv_world490d.dll	C:\opencv\build\x64\vc16\bin
opencv_world490.dll	C:\opencv\build\x64\vc16\bin
opencv_videoio_ffmpeg490_64.dll	C:\opencv\build\x64\vc16\bin

---

7. Add cuDNN DLL to Darknet

Copy cudnn64_7.dll to build\darknet\x64.

---

8. Download Required Files

1. YOLOv4 Weights:
   Download yolov4.conv.137 from YOLOv4 Weights and place it in build\darknet\x64.

2. Cityscapes2Foggy Dataset:
   Download Cityscaples2Foggy.zip from Google Drive and extract it to build\darknet\x64\data.

---

Usage

Navigate to the x64 directory:

cd Project_dir\darknet\build\darknet\x64

Training MS-DAYOLO

To start training, run:

darknet detector train data/c2f.data cfg/ms-dayolo.cfg yolov4.conv.137 -dont_show -map -da

Evaluation

After training, evaluate the model with:

darknet detector map data/c2f.data cfg/ms-dayolo.cfg backup/ms-dayolo_best.weights

Prediction

After evaluation, predict the model with:

darknet detector demo data/c2f.data cfg/ms-dayolo.cfg backup/ms-dayolo_best.weights <video_file>

---

Additional Information

For further details, refer to the original repository: GitHub - Mazin-Hnewa/MS-DAYOLO

Compiling can be more tedious, but following these steps with the compiled one will save you much trouble as long as you set the same environment to mine

---