Car Recognition

This repository is to do car recognition by fine-tuning ResNet-50 with Cars Dataset from Stanford.

Dependencies

• Numpy 1.16.1
• Tensorflow 1.13.1
• Keras 2.0.0
• OpenCV 4.2.0

Dataset

We use the Cars Dataset, which contains 16,185 images of 196 classes of cars. The data is split into 8,144 training images and 8,041 testing images, where each class has been split roughly in a 80-20 split.

You can get it from Cars Dataset:

$ cd Car-Recognition
$ wget http://imagenet.stanford.edu/internal/car196/cars_train.tgz
$ wget http://imagenet.stanford.edu/internal/car196/cars_test.tgz
$ wget --no-check-certificate https://ai.stanford.edu/~jkrause/cars/car_devkit.tgz

ImageNet Pretrained Models

Download ResNet-50 into models folder.

Usage

Data Pre-processing

Extract 8,144 training images, and split them by 80:20 rule (6,515 for training, 1,629 for validation):

$ python pre_process.py

Train

$ python train.py

If you want to visualize during training, run in your terminal:

$ tensorboard --logdir path_to_current_dir/logs

Python version must be atleast 3.6.8 to run this command in your cmd or anaconda promopt or miniconda prompt

Analysis

Update "model_weights_path" in "utils.py" with your best model, and use 1,629 validation images for result analysis:

$ python analyze.py

Validation acc:

97.54%

Confusion matrix:

Test

$ python test.py

Submit predictions of test data set (8,041 testing images) at Cars Dataset, evaluation result:

Test acc:

87.60%

Demo

Download pre-trained model into "models" folder then run:

$ python demo.py

If no argument, a sample image is used:

$ python demo.py
class_name: Lamborghini Reventon Coupe 2008
prob: 0.9999994

GUI

$ python gui.py  

Select a directory of images of cars. It will create bounding box around a car and will show class name and probability score: