zhaofan0622/TF-SSD-VehicleDetection

Vehicle Detection with SSD algorithm and Tensorflow Tools

TF-SSD-VehicleDetection

Vehicle Detection with SSD algorithm and Tensorflow

Notice:

• (1) TensorFlow Version: 1.4; Python Version: 3.6 (Anaconda3);
• (2) My trained models and logs can be download from Google Drive https://drive.google.com/open?id=1_IpPGwND0D0HPCJ9zNAKAInv5J5GaB2g
• (3) Since images in KITTI are 1242x375 but my trained model is SSD300, the detection performance is poor. To improve the performance, you can try to train a new SSD model with larger input images.

How to run this project?

(1) Download and unzip KITTI dataset

Specifically, you only need download 3 files:

• Download left color images of object data set (12 GB)

• Download training labels of object data set (5 MB)

• Download object development kit (1 MB)

(2) Run trainer.py for training you model

please adjust some configurations yourself, such as learning rate, batchsize and so on.

(3) Run tester.py for testing you model

P.S. my chinese blog is https://blog.csdn.net/shuzfan/article/details/79034555