CSE515_MWDB_Project/Phase 1

Phase 1

Getting started: PyTorch, Caltech101 dataset, ResNet50 and similarity measures

• Refer phase1_project23.pdf for problem description
• For task 3, the best distance measures seem to be:
  • Color moments - Pearson (faces especially)
  • Histogram of oriented gradients (HOG) - Cosine similarity
  • ResNet50 (avgpool, layer3, fc) - unsatisfactory results for all, simply used euclidean

Requirements and dependencies

• Requires MongoDB server (local or otherwise)
• Install packages from requirements.txt

Task 1 - task_1.ipynb

After installing, run all cells in the notebook. There will be a prompt to give input for image ID. Range is 0 to 8677

Task 2 - task_2.ipynb

Dataset processing and storage to database. Ensure MongoDB server is running, modify connection URI as needed if running on Atlas

Task 3 - task_3.ipynb

Execute all cells till before the "Target images" markdown cell. There will be four prompts to give input for:

• Image ID: integer, 0 to 8677
• No. of similar images needed, k: positive integer
• Feature model - one of ["cm", "hog", "avgpool", "layer3", "fc"]
  • Note: only hog is applicable for all images. Others cannot be used for grayscale images
• Similarity/distance measure - one of ["euclidean", "cosine", "pearson"]