readme

Author: lindsey98

Influence_function/influence_function.py

@@ -1,15 +1,7 @@
-import faiss
-import numpy as np
-import torch
-import torchvision
 import loss
 from networks import Feat_resnet50_max_n, bninception
 from utils import get_wrong_indices
 import torch.nn as nn
-import os
-import matplotlib.pyplot as plt
-from torchvision.transforms.functional import normalize, resize, to_pil_image
-from torchvision.io.image import read_image
 from Influence_function.EIF_utils import *
 from Influence_function.IF_utils import *
 import pickle

README.md

@@ -3,22 +3,22 @@ Debugging and Explaining Metric Learning Approach: An Influence Function Perspec
 ==============================================================================
 
 ## Introduction
-Deep metric learning (DML) learns a generalizable embedding space of a dataset,
-where semantically similar samples are mapped closer.
+Deep metric learning (DML) learns a generalizable embedding space of a dataset, where semantically similar samples are mapped closer.
 Recently, the record-breaking methodologies have been generally evolving from pairwise-based approaches to proxy-based approaches.
 However, many recent works begin to achieve only marginal improvements on the classical datasets.
 Thus, the explanation approaches of DML are in need for understanding
-**why the trained model can confuse the dissimilar samples and cannot recognize the similar samples**.
+**why the trained model can confuse the dissimilar samples?**.
 
-To answer the above question, we conduct extensive experiments by running 2 comparable state-of-the-art DML approaches.
-The observation leads us to design an influence function based explanation framework to investigate the existing datasets, consisting of:
+The question motivates us to design an influence function based explanation framework to investigate the existing datasets, consisting of:
 - [x] Scalable training-sample attribution:
     - We propose empirical influence function to identify what training samples contribute to the generalization errors, and quantify how much contribution they make to the errors.
 - [x] Dataset relabelling recommendation:
     - We further aim to identify the potentially ``buggy'' training samples with mistaken labels and generate their relabelling recommendation.
 
 ## Requirements
-Install torch, torchvision compatible with your CUDA, see here: https://pytorch.org/get-started/previous-versions/
+- Step 1: Install torch, torchvision compatible with your CUDA, see here: [https://pytorch.org/get-started/previous-versions/](https://pytorch.org/get-started/previous-versions/)
+- Step 2: Install faiss compatible with your CUDA, see here: [https://github.com/facebookresearch/faiss/blob/main/INSTALL.md](https://github.com/facebookresearch/faiss/blob/main/INSTALL.md)
+- Step 3: 
 ```
 pip install -r requirements.txt
 ```
@@ -37,6 +37,22 @@ Put them under mnt/datasets/
 - We use the same hyperparameters specified in [Proxy-NCA++](https://github.com/euwern/proxynca_pp), except for In-Shop we reduce the batch size to 32 due to the limit of our GPU resources.
 
 ## Project Structure
+```
+|__ config/: training config json files
+|__ dataset/: define dataloader
+|__ mnt/datasets/
+   |__ CARS_196/
+   |__ CUB200_2011/
+   |__ inshop/
+|__ evaluation/: evaluation script for recall@k, NMI etc.
+|__ experiments/: scripts for experiments
+|__ Influence_function/: implementation of IF and EIF
+|__ train.py: normal training script
+|__ train_noisy_data.py: noisy data trianing script
+|__ train_sample_reweight.py: re-weighted training script
+```
+
+## Instructions
 - Training the original models
   - Training the DML models with Proxy-NCA++ loss or with SoftTriple loss
 ```
@@ -81,10 +97,6 @@ python train_noisydata.py --dataset [cub_noisy|cars_noisy|inshop_noisy] \
 
      See experiments/sample_recommendation_evaluation.py
 
-- Implementation of EIF
-
-      See Influence_function/influence_function.py
-
 ## Results
 - All trained models: https://drive.google.com/drive/folders/1uzy3J78iwKZMCx_k5yESDLbcLl9RADDb?usp=sharing
 - For the detailed statistics of Table 1, please see https://docs.google.com/spreadsheets/d/1f4OXVLO2Mu2CHrBVm72a2ztTHx5nNG92dczTNNw7io4/edit?usp=sharing

loss.py

@@ -3,10 +3,6 @@
 from similarity import pairwise_distance
 import torch.nn.functional as F
 import sklearn.preprocessing
-import logging
-from scipy.optimize import linear_sum_assignment
-import utils
-from tqdm import tqdm
 
 def masked_softmax(A, dim, t=1.0):
     '''

requirements.txt

@@ -2,7 +2,9 @@ scikit-learn
 matplotlib
 scipy
 tqdm
-faiss-cpu
 h5py
 torchcam
-rembg
+rembg
+numpy
+argparse
+Pillow

train_sample_reweight.py

@@ -6,13 +6,13 @@
 import torch
 import numpy as np
 import matplotlib
-matplotlib.use('agg', force=True)
 import time
 import argparse
 import json
 from tqdm import tqdm
 from torch.utils.data import Dataset, DataLoader
 import random
+matplotlib.use('agg', force=True)
 
 os.environ["CUDA_VISIBLE_DEVICES"]="1, 0"