This is the project for the deep learning course at University of Trento (Italy).
This work was done with two other students friends of mine.
The project exploits video surveillance data, more precisely the Market1501 dataset.
This work consists of two main tasks:
- Design and implementation of a multi task classifier for attribute classification (age, gender and clothing).
- Person re-identification, given a query image the goal is to retrieve all the images of the same person from the dataset.
This repo contains:
- The colab notebooks.
- The report.
You can open the colab notebooks from the links inside the papers.
The notebooks are auto-contained (they will download the dataset and install the required packages on their own), so you can just run "execute all" inside google colab. Remember to select the GPU environment.
-
FINAL_TASK_1 : trains a NN based on resnet18 to classify people attributes, then it generates the .csv (running time: 5-6 mins).
-
FINAL_TASK_2 : trains Bruna (Based on Residuals hUge Architecture) for 42 epochs (50-60 mins) then saves the weights locally.
-
TASK2_INFERENCE : download our pretrained models (for both task 1&2) and uses them to retrieve people given a query (45 mins).
-
Others/SplitDataset : shows some stats on the split of the dataset.
-
other resources: Same Resources But In GDrive