CAR DAMAGE INTENSITY DATASET
The dataset contains 3 folders containing images which describes the intensity of damages as minor , moderate and severe accordingly. The task was to analyse and explore the different feature extraction techniques along with classifiers to check upon the performance of the system. Hereby I worked with 3 different types of approaches to tackle the task
APPROACH 1: Using Feature Extraction:
APPROACH 2: Using Convolutional Neural Networks (CNN).
APPROACH 3: Using PreTrained models approach (best one)
For detailed description of the methods used, please refer to the pdf attached with the project
Conclusion: This experiments actually concludes that Pretrained model for CNNs performs better than the conventional feature extraction techniques for any type of complex classification task and also reduce the time consumed to construct a deep neural net from scratch.