Performance Analysis of Person Re Identification Using Global Local image
Implementation Plan:
——————————-
Step 1: Initially, load the input images from three visible-infrared datasets.
Step 2: Next, Perform the Data Preprocessing by this process includes data augmentation, resizing, and normalization to prepare the data for training.
Step 3: Next , We implement a novel Global-local Transformer based model for Visible-infrared person reidentification.
Step 4: Next we Split the dataset into training and validation sets and implement loss functions like triplet loss or contrastive loss or Modality-
Aware Enhancement (MAE) loss, softmax loss, which are commonly used for Re-ID.
Step 5: Next we Train the model using the training dataset while validating on the validation set and test the model on the test dataset.
Step 6: Finally, need to evaluate the model using the following performance metrics, Such as,
6.1 Cumulative Matching Characteristics Curve (Rank-1, Rank-5, Rank-10 and Rank-20)
6.2: Mean Average precision (mAP)
6.3: Mean Inverse Negative Penalty(mINP)
======================================================================================================================================
For your kind consideration I have sent you three paper links, you can follow these.
https://ieeexplore.ieee.org/abstract/document/9963608
https://ieeexplore.ieee.org/abstract/document/9725265
https://ieeexplore.ieee.org/abstract/document/10130375
If you need help about code you can check the following link
https://github.com/alehdaghi/Cross-Modal-Re-ID-via-LUPI
