cchamikara / AI-for-S.E.A-Computer-Vision-Challenge

This repository is a submission for the AI for S.E.A Computer Vision Challenge

AI-for-S.E.A-Computer-Vision-Challenge

This repository is a submission for the Grab AI for S.E.A Computer Vision Challenge.

Getting Started

Download the Data

We use a preprocessed dataset containing the Stanford Car dataset by classes folder. The dataset can be found here: https://www.kaggle.com/jutrera/stanford-car-dataset-by-classes-folder

the dataset includes:

1. images of cars by classes folder (car_data.zip)
2. annotation for the train dataset (anno_train.csv)
3. annotation for the test dataset (anno_test.csv)
4. names of the cars (names.csv)

We prepocess the dataset by adding transformations and normalization to both the training and test set.

Dependencies

We use the standard PyTorch libraries as well as supporting libraries

Model

The model we use is a pretrained ResNet50 model from torchvision.models library.

Model Reproducibility

We train the model on a single Tesla V100 GPU with a 89% accuracy on the test dataset. We implement an automated model saving that can be used later to reproduce the model performance. The instructions are included in the Stanford_Car_ResNet50.ipynb file.