resolved conflict

visionjo · Sep 11, 2019 · 9fb2780 · 9fb2780
2 parents 072c000 + d5d57c0
commit 9fb2780
Show file tree

Hide file tree

Showing 53 changed files with 774 additions and 558 deletions.
diff --git a/README.md b/README.md
@@ -173,7 +173,7 @@ By downloading the image data you agree to the following terms:
 
 
 ## Authors
-* **Joseph Robinson** - [Github](https://github.com/huskyjo) - [web](http://www.jrobsvision.com)
+* **Joseph Robinson** - [Github](https://github.com/visionjo) - [web](http://www.jrobsvision.com)
 
 ######
 ### Contact

diff --git a/data/.gitkeep b/data/.gitkeep
diff --git a/data/external/.gitkeep b/data/external/.gitkeep
diff --git a/data/interim/.gitkeep b/data/interim/.gitkeep
diff --git a/data/processed/.gitkeep b/data/processed/.gitkeep
diff --git a/data/raw/.gitkeep b/data/raw/.gitkeep
diff --git a/notebooks/bin_members.ipynb b/notebooks/bin_members.ipynb
diff --git a/notebooks/explore_fiw.ipynb b/notebooks/explore_fiw.ipynb
diff --git a/references/.DS_Store b/references/.DS_Store
diff --git a/references/ACM-MM-poster-2016.pdf b/references/ACM-MM-poster-2016.pdf
diff --git a/references/fiw-ne-cv-workshop-2016.pdf b/references/fiw-ne-cv-workshop-2016.pdf
diff --git a/references/fiw-poster-2017.pdf b/references/fiw-poster-2017.pdf
diff --git a/references/fiw_acm-mm-2016.pdf b/references/fiw_acm-mm-2016.pdf
diff --git a/references/fiw_mdml_FG2017.pdf b/references/fiw_mdml_FG2017.pdf
diff --git a/requirements.txt b/requirements.txt
@@ -1,3 +1,7 @@
+dlib
+torchvision
+matplotlib
+torch
 Pillow
 scipy
 scikit-learn

diff --git a/setup.py b/setup.py
@@ -1,23 +1,29 @@
 # -*- coding: utf-8 -*-
 
 from setuptools import setup, find_packages
+import setuptools
 
-
-with open('README.md') as f:
+with open('README.md', 'r') as f:
     readme = f.read()
 
-with open('LICENSE') as f:
+with open('LICENSE', 'r') as f:
     license = f.read()
 
 setup(
-    name='FIW_KRT',
-    version='1.0.0',
+    name='fiwtools',
+    version='0.1.0',
     description='Families In the WIld: A Kinship Recogntion Toolbox.',
     long_description=readme,
     author='Joseph Robinson',
     author_email='robinson.jo@husky.neu.edu',
-    url='https://github.com/huskyjo/FIW_KRT',
+    url='https://github.com/visionjo/FIW_KRT',
+    packages=setuptools.find_packages(),
     license=license,
-    packages=find_packages(exclude=('tests', 'docs'))
+    # packages=find_packages(exclude=('tests', 'docs'))
+    classifiers=[
+        "Programming Language :: Python :: 3",
+        "License :: OSI Approved :: MIT License",
+        "Operating System :: OS Independent",
+    ],
 )
 
diff --git a/sphereface_rfiw_baseline/data_loader.py b/sphereface_rfiw_baseline/data_loader.py
@@ -1,3 +1,4 @@
+<<<<<<< HEAD
 from torch.utils import data
 from torchvision import transforms as T
 from torchvision.datasets import ImageFolder
@@ -18,10 +19,38 @@ def __init__(self, root_dir, labels_path, n_classes, transform):
         self.transform = transform
         self.train_dataset = []
         self.preprocess()
+=======
+from torch.utils.data import Dataset, DataLoader
+from torchvision import transforms as T
+from PIL import Image
+import os
+import csv
+
+
+class FIW_DBBase(Dataset):
+    def __init__(self, root_dir, labels_path, n_classes=300, transform=None):
+        self.root_dir = root_dir
+        self.labels_path = labels_path
+        self.n_classes = n_classes
+        self.transform = transform
+        self.pairs = []
+
+    def __len__(self):
+        return len(self.pairs)
+
+    def __getitem__(self, idx):
+        pass
+
+
+
+class FIW_Train(FIW_DBBase):
+    """Dataset class for FIW Training Set"""
+>>>>>>> d5d57c0ffdab6a2eac16d8809ae66bd6ab8f5f19
 
     def preprocess(self):
         """Process the labels file"""
         lines = [line.rstrip() for line in open(self.labels_path, 'r')]
+<<<<<<< HEAD
 
         for l in lines:
             spt = l.split()
@@ -65,10 +94,45 @@ def __getitem__(self, index):
         path_a, path_b, label = self.val_dataset[index]
         img_a = self.transform(Image.open(os.path.join(self.base_dir, path_a)))
         img_b = self.transform(Image.open(os.path.join(self.base_dir, path_b)))
+=======
+        for l in lines:
+            spt = l.split()
+            fname = spt[0]
+            label = int(spt[1])
+            # label_vec = [1 if i == label else 0 for i in range(self.n_classes)]
+            self.pairs.append([fname, label])
+
+    def __getitem__(self, index):
+        """Return an image"""
+        filename, label = self.pairs[index]
+        impath = os.path.join(self.root_dir, filename)
+        image = Image.open(impath)
+        return self.transform(image), label
+
+
+
+class FIW_Val(FIW_DBBase):
+    """Dataset class for FIW Validation Set"""
+
+    def preprocess(self):
+        """Process the pair CSVs"""
+        with open(self.labels_path, 'r') as f:
+            re = csv.reader(f)
+            lines = list(re)
+
+        self.pairs = [(l[2], l[3], bool(int(l[1]))) for l in lines]
+
+    def __getitem__(self, index):
+        """Return a pair"""
+        path_a, path_b, label = self.pairs[index]
+        img_a = self.transform(Image.open(os.path.join(self.root_dir, path_a)))
+        img_b = self.transform(Image.open(os.path.join(self.root_dir, path_b)))
+>>>>>>> d5d57c0ffdab6a2eac16d8809ae66bd6ab8f5f19
         return (img_a, img_b), label
 
     def __len__(self):
         """Return the number of images."""
+<<<<<<< HEAD
         return len(self.val_dataset)
 
 def get_train_loader(image_dir, labels_path='train/train.label', n_classes=300, image_size=(112, 96), batch_size=16, num_workers=1):
@@ -86,11 +150,32 @@ def get_train_loader(image_dir, labels_path='train/train.label', n_classes=300,
     dataset = FIW_Train(image_dir, labels_path, n_classes, transform)
 
     data_loader = data.DataLoader(dataset=dataset,
+=======
+        return len(self.pairs)
+
+
+
+
+
+def get_train_loader(image_dir, labels_path='train/train.label', n_classes=300, image_size=(112, 96), batch_size=16,
+                     num_workers=1):
+    """Build and return a data loader for the training set."""
+    transform = T.Compose([T.RandomHorizontalFlip(),
+                           T.Resize(image_size),
+                           T.ToTensor(),
+                           T.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))
+                           ])
+
+    dataset = FIW_Train(image_dir, labels_path, n_classes=n_classes, transform=transform)
+    dataset.preprocess()
+    data_loader = DataLoader(dataset=dataset,
+>>>>>>> d5d57c0ffdab6a2eac16d8809ae66bd6ab8f5f19
                                   batch_size=batch_size,
                                   shuffle=True,
                                   num_workers=num_workers)
     return data_loader
 
+<<<<<<< HEAD
 def get_val_loader(base_dir, csv_path, image_size=(112, 96), batch_size=128, num_workers=1):
     """Build and return a data loader for a split in the validation set."""
     transform = []
@@ -103,6 +188,19 @@ def get_val_loader(base_dir, csv_path, image_size=(112, 96), batch_size=128, num
     dataset = FIW_Val(base_dir, csv_path, transform)
 
     data_loader = data.DataLoader(dataset=dataset,
+=======
+
+def get_val_loader(base_dir, csv_path, image_size=(112, 96), batch_size=128, num_workers=1):
+    """Build and return a data loader for a split in the validation set."""
+    transform = T.Compose([T.Resize(image_size),
+                           T.ToTensor(),
+                           T.Normalize(mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5))
+                           ])
+
+    dataset = FIW_Val(base_dir, csv_path, transform=transform)
+    dataset.preprocess()
+    data_loader = DataLoader(dataset=dataset,
+>>>>>>> d5d57c0ffdab6a2eac16d8809ae66bd6ab8f5f19
                                   batch_size=batch_size,
                                   shuffle=False,
                                   num_workers=num_workers)

diff --git a/sphereface_rfiw_baseline/evaluate_rfiw.py b/sphereface_rfiw_baseline/evaluate_rfiw.py
@@ -0,0 +1,114 @@
+import numpy as np
+import argparse
+import net_sphere
+import os
+from torchtools import cuda, TorchTools, Tensor
+from data_loader import get_val_loader
+from sklearn.metrics import roc_curve, auc
+import matplotlib.pyplot as plt
+from torch.autograd import Variable
+from fiwtools.utils.io import sys_home
+import torch.nn.functional as F
+do_plot = True
+
+
+def initialize_roc_plot(ax, lw=2):
+    ax.plot([0, 1], [0, 1], color='navy', lw=lw, linestyle='--')
+    plt.xlim([0.0, 1.0])
+    plt.ylim([0.0, 1.05])
+    # ax.xlabel('False Positive Rate')
+    # ax.ylabel('True Positive Rate')
+    # ax.title('Receiver operating characteristic example')
+    # ax.legend(loc="lower right")
+    return ax
+
+
+def generate_roc(ax, tp_array, fp_array, roc_auc, color='darkred', lw=2, label='ROC curve (area = %0.2f)', init=False, fname=''):
+
+
+    # if init:
+    #     initialize_roc_plot(lw)
+    ax.plot(fp_array, tp_array, color=color, lw=lw, label=label % roc_auc)
+    # if len(fname) > 0:
+    #     plt.savefig(fname)
+def validate(net, data_loader, ax):
+    print('Begin validation')
+    net.eval()
+    y_labels = []
+    distances = []
+    ii=0
+    for pairs, labels in iter(data_loader):
+        if ii > 10:
+            break
+        img_a = Variable(pairs[0]).type(Tensor)
+        img_b = Variable(pairs[1]).type(Tensor)
+
+        _, embs_a = net(img_a)
+        _, embs_b = net(img_b)
+
+        embs_a = embs_a.data
+        embs_b = embs_b.data
+        cos_dis = F.cosine_similarity(embs_a, embs_b)
+        distances += list(cos_dis.data.cpu().numpy())
+
+        y_labels += list(labels.numpy())
+        # ii += 1
+
+    dist_array = np.array(distances)
+    y_array = np.array(y_labels)
+
+    fpr, tpr, thresh = roc_curve(y_array, dist_array)
+    roc_auc = auc(fpr, tpr)
+
+    if do_plot:
+        initialize_roc_plot(ax)
+        generate_roc(ax, tpr, fpr, roc_auc, color='darkred')
+        # fpr_micro, tpr_micro, _ = roc_curve(y_array.ravel(), dist_array.ravel())
+        # roc_auc_micro = auc(fpr_micro, tpr_micro)
+        #
+        # generate_roc(tpr_micro, fpr_micro, roc_auc_micro, color='darkorange')
+
+        # plt.show()
+
+    return roc_auc
+
+do_types = np.linspace(0, 6, 7).astype(np.uint8)
+types = ['bb', 'ss', 'sibs', 'fd', 'fs', 'md', 'ms']
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description='FIW Sphereface Baseline')
+    parser.add_argument('--type', '-t', default='bb', type=str, help='relationship type (None processes entire directory)')
+    parser.add_argument('--batch_size', default=32, type=int, help='training batch size')
+    parser.add_argument('--modelpath', default='finetuned/checkpoint.pth.tar', type=str,
+                        help='the pretrained model to point to')
+    parser.add_argument('--label_dir', '-l', type=str, default=sys_home() + '/datasets/FIW/RFIW/val/pairs/',
+                            help='Root directory of data (assumed to containing pairs list labels)')
+    parser.add_argument('--data_dir', '-d', type=str, default=sys_home() + '/datasets/FIW/RFIW/val/',
+                        help='Root directory of data (assumed to contain valdata)')
+
+    args = parser.parse_args()
+
+    net = net_sphere.sphere20a(classnum=300)
+
+    if cuda:
+        net.cuda()
+
+    epoch, bess_acc = TorchTools.load_checkpoint(net, f_weights=args.modelpath)
+
+    ncols = int(np.ceil(len(do_types) / 2))
+    nrows = 2
+    f, axes = plt.subplots(nrows, ncols, sharex='all', sharey='all')
+
+    for i, id in enumerate(do_types):
+        if i < ncols:
+            ax = axes[0, i]
+        else:
+            ax = axes[1, i - ncols]
+        csv_file = os.path.join(args.label_dir, types[id] + '_val.csv')
+        loader = get_val_loader(args.data_dir, csv_file)
+        # f.subplot()
+        auc_score = validate(net, loader, ax)
+
+        print('{} pairs: {} (auc)'.format(types[id], auc_score))
+
+    plt.savefig('roc.png')