Ajithjaas / panorama_stitching

Phase 1 - Traditional Method

Steps in Traditional Method:

- Corner Detection
- Adaptive Non-Maximal Suppression (ANMS)
- Feature Descriptor
- Feature Matching
- RANSAC for Outlier Rejection
- Blending Images

Navigate the below folder:

cd ../Phase1/Code/

Run the follwing code:

python Wrapper.py

Phase 2 - Estimation of Homography using Deep Learning

Instructions to run the code

Please make sure you are in the same directory as the code files in order to execute them in the command prompt. If not, use cd/../loc to navigate to the location of the Python files.

Navigate the below folder:

cd ../Phase2/Code/

Creating the data

The following command must be entered into the command prompt to create Training data:

python Wrapper.py --NumFeatures 64 --Data Train

The following command must be entered into the command prompt to create Testing Data:

python Wrapper.py --NumFeatures 64 --Data Test

Training Model

The following command must be entered into the command prompt to run training:

python Train.py --NumFeatures 64 --ModelType Sup

Based on the Model to run, the --ModelType can either Sup for supervised network and Unsup for Unsupervised Network. In case you need more information regarding the parse arguments for a particular file, you can run the following code:

python Train.py -h

The following command must be entered into the command prompt to run testing:

 python Test.py --NumFeatures 64