-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Marsha Gómez edited this page Jan 14, 2021
·
2 revisions
Welcome to the DNN-Sketches-image-analysis wiki!
- Implement index 'LSH' to allow fast similarity search on deep features and create an image search engine on top of it
- Use the pre-trained Deep Neural Network 'Inception' to extract features from the dataset 'Sketches' and the distractor 'MirFlickr'
- Index the extracted features using your search engine
- Measure the retrieval performance of the image search engine
- Fine tune the 'Inception' for 'Sketches'
- Use the fine-tuned 'Inception' to extract features from 'Sketches' and 'MirFlickr'
- Index the new extracted features using your search engine
- Measure the retrieval performance of the image search engine using the new features (to this end, use the mAP we have seen in Lab Session 6 on a fixed set of queries, of your choice, taken from your test set)
- Compare the performance of the two features
- Optional:
- Build a web based user interface for your web search engine