Extract Feat

ExtractFeat/datasets/__init__.py

@@ -0,0 +1,4 @@
+# --------------------------------------------------------
+# Domain Adaptation
+# Written by VARMS
+# --------------------------------------------------------

ExtractFeat/datasets/data_loader.py

@@ -0,0 +1,50 @@
+import os
+import datasets.list_loader as list_loader
+import torch
+from torchvision import transforms
+import lib.utils as utils
+from lib.config import cfg
+import torch.distributed as dist
+from os.path import join as join
+
+
+def get_transform():
+    trans = []
+
+    if cfg.AUG.RESIZE[0] > 0 and cfg.AUG.RESIZE[1] > 0:
+        trans.append(transforms.Resize(cfg.AUG.RESIZE))
+    if cfg.AUG.V_FLIP > 0:
+        trans.append(transforms.RandomVerticalFlip(p=cfg.AUG.V_FLIP))
+    if cfg.AUG.H_FLIP > 0:
+        trans.append(transforms.RandomHorizontalFlip(p=cfg.AUG.H_FLIP))
+    if cfg.AUG.ROTATION > 0:
+        trans.append(transforms.RandomRotation(cfg.AUG.ROTATION, expand=False))
+    if cfg.AUG.BRIGHTNESS > 0 or cfg.AUG.CONTRAST > 0 or cfg.AUG.SATURATION > 0 or cfg.AUG.HUE > 0:
+        trans.append(transforms.ColorJitter(brightness=cfg.AUG.BRIGHTNESS, 
+            contrast=cfg.AUG.CONTRAST, saturation=cfg.AUG.SATURATION, hue=cfg.AUG.HUE))
+    if cfg.AUG.RND_CROP[0] > 0 and cfg.AUG.RND_CROP[1] > 0:
+        trans.append(transforms.RandomCrop(cfg.AUG.RND_CROP))
+
+    trans.append(transforms.ToTensor())
+    trans.append(transforms.Normalize(cfg.MEAN, cfg.STD))
+    return trans
+
+
+def load_test(target_root, test_label, use_mirror = False):
+    if use_mirror:
+        transform = transforms.Compose([
+            transforms.RandomHorizontalFlip(p=1.0),
+            transforms.Resize(cfg.AUG.TEST_CROP),
+            transforms.ToTensor(),
+            transforms.Normalize(cfg.MEAN, cfg.STD)
+        ])
+    else:
+        transform = transforms.Compose([
+            transforms.Resize(cfg.AUG.TEST_CROP),
+            transforms.ToTensor(),
+            transforms.Normalize(cfg.MEAN, cfg.STD)
+        ])
+    image_set = list_loader.ListLoader(target_root, test_label, transform)
+    loader = torch.utils.data.DataLoader(image_set, batch_size=cfg.TEST.BATCH_SIZE,
+        shuffle=False, num_workers=cfg.DATA_LOADER.NUM_WORKERS)
+    return loader

ExtractFeat/datasets/list_loader.py

@@ -0,0 +1,43 @@
+import torch
+import torch.utils.data as data
+from PIL import Image
+import os
+import random
+import numpy as np
+from lib.config import cfg
+
+
+def default_loader(path):
+    return Image.open(path).convert('RGB')
+
+
+def make_dataset(root, list_path):
+    images = []
+    listtxt = open(list_path)
+    for line in listtxt:
+        data = line.strip().split(' ')
+        path = os.path.join(root, data[0])
+        label = int(data[1])
+        item = (path, label)
+        images.append(item)
+    return images
+
+
+class ListLoader(data.Dataset):
+    def __init__(self, root, list_path, transform=None, loader=default_loader):
+        imgs = make_dataset(root, list_path)
+        self.root = root
+        self.imgs = imgs
+        self.transform = transform
+        self.loader = loader
+
+    def __getitem__(self, index):
+        path, target = self.imgs[index]
+        img = self.loader(path)
+
+        if self.transform is not None:
+            img = self.transform(img)
+        return img, target
+
+    def __len__(self):
+        return len(self.imgs)

ExtractFeat/experiments/phase1/clipart/efficientnet_b4/efficientnet_b4_clipart/config.yml

@@ -0,0 +1,43 @@
+LOGGER_NAME: 'log'
+SEED: 1546884941.160048
+############################ TRAIN ############################
+TRAIN:
+  FIX_TWO: True
+
+############################ TEST ############################
+TEST:
+  BATCH_SIZE: 32
+
+############################ DATA_LOADER ############################
+DATA_LOADER:
+  NUM_WORKERS: 4
+  SHUFFLE: True
+  DATA_ROOT: '../dataset/visda2019'
+  LIST: 'list'
+  FOLDER: 'clipart'
+
+############################ AUG ############################
+AUG:
+  TEST_CROP: [320, 320]
+  #TEST_CROP: [288, 288]
+  RESIZE: [0, 0]
+  RND_CROP: [0, 0]
+
+  V_FLIP: 0.0
+  H_FLIP: 0.5
+  ROTATION: 0.0
+  BRIGHTNESS: 0.0
+  CONTRAST: 0.0
+  SATURATION: 0.0
+  HUE: 0.0
+  MULTI_CROP_SIZE: 0
+
+MEAN: [0.485, 0.456, 0.406]
+STD:  [0.229, 0.224, 0.225]
+
+############################ MODEL ############################
+MODEL:
+  CLASS_NUM: 345
+  NET: 'efficientnet_b4'
+  IN_DIM: 1792
+  EMBED_DIM: 1000

ExtractFeat/experiments/phase1/clipart/efficientnet_b4/efficientnet_b4_infograph/config.yml

@@ -0,0 +1,43 @@
+LOGGER_NAME: 'log'
+SEED: 1546884941.160048
+############################ TRAIN ############################
+TRAIN:
+  FIX_TWO: True
+
+############################ TEST ############################
+TEST:
+  BATCH_SIZE: 32
+
+############################ DATA_LOADER ############################
+DATA_LOADER:
+  NUM_WORKERS: 4
+  SHUFFLE: True
+  DATA_ROOT: '../dataset/visda2019'
+  LIST: 'list'
+  FOLDER: 'infograph'
+
+############################ AUG ############################
+AUG:
+  TEST_CROP: [320, 320]
+  #TEST_CROP: [288, 288]
+  RESIZE: [0, 0]
+  RND_CROP: [0, 0]
+
+  V_FLIP: 0.0
+  H_FLIP: 0.5
+  ROTATION: 0.0
+  BRIGHTNESS: 0.0
+  CONTRAST: 0.0
+  SATURATION: 0.0
+  HUE: 0.0
+  MULTI_CROP_SIZE: 0
+
+MEAN: [0.485, 0.456, 0.406]
+STD:  [0.229, 0.224, 0.225]
+
+############################ MODEL ############################
+MODEL:
+  CLASS_NUM: 345
+  NET: 'efficientnet_b4'
+  IN_DIM: 1792
+  EMBED_DIM: 1000

ExtractFeat/experiments/phase1/clipart/efficientnet_b4/efficientnet_b4_quickdraw/config.yml

@@ -0,0 +1,43 @@
+LOGGER_NAME: 'log'
+SEED: 1546884941.160048
+############################ TRAIN ############################
+TRAIN:
+  FIX_TWO: True
+
+############################ TEST ############################
+TEST:
+  BATCH_SIZE: 32
+
+############################ DATA_LOADER ############################
+DATA_LOADER:
+  NUM_WORKERS: 4
+  SHUFFLE: True
+  DATA_ROOT: '../dataset/visda2019'
+  LIST: 'list'
+  FOLDER: 'quickdraw'
+
+############################ AUG ############################
+AUG:
+  TEST_CROP: [320, 320]
+  #TEST_CROP: [288, 288]
+  RESIZE: [0, 0]
+  RND_CROP: [0, 0]
+
+  V_FLIP: 0.0
+  H_FLIP: 0.5
+  ROTATION: 0.0
+  BRIGHTNESS: 0.0
+  CONTRAST: 0.0
+  SATURATION: 0.0
+  HUE: 0.0
+  MULTI_CROP_SIZE: 0
+
+MEAN: [0.485, 0.456, 0.406]
+STD:  [0.229, 0.224, 0.225]
+
+############################ MODEL ############################
+MODEL:
+  CLASS_NUM: 345
+  NET: 'efficientnet_b4'
+  IN_DIM: 1792
+  EMBED_DIM: 1000

ExtractFeat/experiments/phase1/clipart/efficientnet_b4/efficientnet_b4_real/config.yml

@@ -0,0 +1,43 @@
+LOGGER_NAME: 'log'
+SEED: 1546884941.160048
+############################ TRAIN ############################
+TRAIN:
+  FIX_TWO: True
+
+############################ TEST ############################
+TEST:
+  BATCH_SIZE: 32
+
+############################ DATA_LOADER ############################
+DATA_LOADER:
+  NUM_WORKERS: 4
+  SHUFFLE: True
+  DATA_ROOT: '../dataset/visda2019'
+  LIST: 'list'
+  FOLDER: 'real'
+
+############################ AUG ############################
+AUG:
+  TEST_CROP: [320, 320]
+  #TEST_CROP: [288, 288]
+  RESIZE: [0, 0]
+  RND_CROP: [0, 0]
+
+  V_FLIP: 0.0
+  H_FLIP: 0.5
+  ROTATION: 0.0
+  BRIGHTNESS: 0.0
+  CONTRAST: 0.0
+  SATURATION: 0.0
+  HUE: 0.0
+  MULTI_CROP_SIZE: 0
+
+MEAN: [0.485, 0.456, 0.406]
+STD:  [0.229, 0.224, 0.225]
+
+############################ MODEL ############################
+MODEL:
+  CLASS_NUM: 345
+  NET: 'efficientnet_b4'
+  IN_DIM: 1792
+  EMBED_DIM: 1000

ExtractFeat/experiments/phase1/clipart/efficientnet_b4/efficientnet_b4_sketch/config.yml

@@ -0,0 +1,43 @@
+LOGGER_NAME: 'log'
+SEED: 1546884941.160048
+############################ TRAIN ############################
+TRAIN:
+  FIX_TWO: True
+
+############################ TEST ############################
+TEST:
+  BATCH_SIZE: 32
+
+############################ DATA_LOADER ############################
+DATA_LOADER:
+  NUM_WORKERS: 4
+  SHUFFLE: True
+  DATA_ROOT: '../dataset/visda2019'
+  LIST: 'list'
+  FOLDER: 'sketch'
+
+############################ AUG ############################
+AUG:
+  TEST_CROP: [320, 320]
+  #TEST_CROP: [288, 288]
+  RESIZE: [0, 0]
+  RND_CROP: [0, 0]
+
+  V_FLIP: 0.0
+  H_FLIP: 0.5
+  ROTATION: 0.0
+  BRIGHTNESS: 0.0
+  CONTRAST: 0.0
+  SATURATION: 0.0
+  HUE: 0.0
+  MULTI_CROP_SIZE: 0
+
+MEAN: [0.485, 0.456, 0.406]
+STD:  [0.229, 0.224, 0.225]
+
+############################ MODEL ############################
+MODEL:
+  CLASS_NUM: 345
+  NET: 'efficientnet_b4'
+  IN_DIM: 1792
+  EMBED_DIM: 1000

ExtractFeat/experiments/phase1/clipart/efficientnet_b5/efficientnet_b5_clipart/config.yml

@@ -0,0 +1,43 @@
+LOGGER_NAME: 'log'
+SEED: 1546884941.160048
+############################ TRAIN ############################
+TRAIN:
+  FIX_TWO: True
+
+############################ TEST ############################
+TEST:
+  BATCH_SIZE: 32
+
+############################ DATA_LOADER ############################
+DATA_LOADER:
+  NUM_WORKERS: 4
+  SHUFFLE: True
+  DATA_ROOT: '../dataset/visda2019'
+  LIST: 'list'
+  FOLDER: 'clipart'
+
+############################ AUG ############################
+AUG:
+  TEST_CROP: [320, 320]
+  #TEST_CROP: [288, 288]
+  RESIZE: [0, 0]
+  RND_CROP: [0, 0]
+
+  V_FLIP: 0.0
+  H_FLIP: 0.5
+  ROTATION: 0.0
+  BRIGHTNESS: 0.0
+  CONTRAST: 0.0
+  SATURATION: 0.0
+  HUE: 0.0
+  MULTI_CROP_SIZE: 0
+
+MEAN: [0.485, 0.456, 0.406]
+STD:  [0.229, 0.224, 0.225]
+
+############################ MODEL ############################
+MODEL:
+  CLASS_NUM: 345
+  NET: 'efficientnet_b5'
+  IN_DIM: 2048
+  EMBED_DIM: 1000