**Learning Features and Parts of Car Using Deep CNN**
Shubham Sarpal
*University of Texas at San Antonio, San Antonio, Texas, USA*
sarpal75@gmail.com
**Project Definition:** Fine-grained recognition refers to a subordinate level of recognition, such as recognizing different species of animals, aircrafts, cars, humans and computers. In this project neural networks (CNNs) is used to learn appearance descriptors and perform unsupervised part discovery to obtain a collection of part detectors.
Datasets incudes of the 10 types of BMW (Bmw-10 series) with 512 images having many view points along bounding boxes and hand-curated. Also 196 car models with 16,185 images. All work will be done on class labels and bounding boxes of the car. It will be implemented by splitting the dataset into two: train and test, images are cropped to their ground truth bounding box this is done for identifying the make and models of cars from various angles and different settings with the added constraint of limited data and time.
1]: Jonathan Krause, Michael Stark, Jia Deng, Li Fei-Fei, 3D Object Representations for Fine-Grained Categorization. 4th IEEE Workshop on 3D Representation and Recognition, at ICCV 2013 (3dRR-13). Sydney, Australia. Dec. 8, 2013.
**Outcome:** In this an object representation is done that detects important parts and describes fine grained appearances using neural networks.
**Dataset:** The image dataset can be found on http://ai.stanford.edu/~jkrause/cars/car_dataset.html. This dataset contains 16,185 image classification pairs of 196 different classes, split into 8,144 training and 8,041 test images. Each classes accordingly order of year, make and model of a vehicle
In [ ]: