Movie Scene segmentation for eluvio
In this challenge,
The task was to look at 64 pkl files which are from 64 movies and identify scene boundaries present in these movies. The paper 'A Local-to-Global Approach to Multi-modal Movie Scene Segmentation' talks extensively about how they have performed an analysis for this task.
I started with the notebook
- Deep_learning_models and then created the notebook machine_learning_models
We are to provide 2 metric values for evaluation, namely mAP and Miou:
- The results obtained with the predictions provided: "mAP": 0.48199051394228226, "Miou": 0.46508357436009745,
- My best model so far which is a neural network : "mAP": 0.26102519658286005, "Miou": 0.38804374907272776,
I would like to thank eluv.io for providing me with this challenge. It was exciting and challenging.