huyquangdao/logodetection

Brand And Logo Detection In Image Using Yolov3

1.Introduction:

This repo is an implementation of our project: "Image and Brand Detection In Image" in course: "Data Mining".

2.Requirements:

• Python: 3.6
• Tensorflow == 1.14.0
• Tensorflow-gpu == 1.14.0

To install this project's package dependencies please type command

pip install -r requirements.txt

3.Running Demo:

Please download the checkpoint of our model in Link

• cd to our project working directory

• type command

    python test_image.py --input_image YOUR_IMAGE_PATH --restore_path CHECKPOINT_PATH
  

Some Result

4.Dataset:

We use Open Logo dataset for training Yolov3 Model

• Training: 75% dataset
• Validate: 25% dataset

Dataset Paper: https://arxiv.org/abs/1807.01964

This dataset you can download in https://qmul-openlogo.github.io/

5.Experiments:

Our project got 42.5 % mAP (mean Average Precision) on validation set

logodetection's Issues

Cannot convert a symbolic Tensor (meshgrid/Size_1:0) to a numpy array.

Hi!
I'm currently trying to use this on code on colab and I'm getting this error when i try to run test_image.py, can you please help