Adjust directory structure

config.py

@@ -9,7 +9,7 @@ NUM_FEATURE_MAP_CHANNEL = 6
 # image_size / 2^5 = 512 / 32 = 16
 FEATURE_MAP_SIZE = 16
 # Thresholds to determine whether an detected point match ground truth.
-SQUARED_DISTANCE_THRESH = 0.0003
+SQUARED_DISTANCE_THRESH = 0.000277778
 DIRECTION_ANGLE_THRESH = 0.5
 
 
@@ -34,11 +34,11 @@ def get_parser_for_training():
                         help="The weights of optimizer.")
     parser.add_argument('--batch_size', type=int, default=24,
                         help="Batch size.")
-    parser.add_argument('--data_loading_workers', type=int, default=24,
+    parser.add_argument('--data_loading_workers', type=int, default=48,
                         help="Number of workers for data loading.")
     parser.add_argument('--num_epochs', type=int, default=100,
                         help="Number of epochs to train for.")
-    parser.add_argument('--lr', type=float, default=1e-3,
+    parser.add_argument('--lr', type=float, default=1e-4,
                         help="The learning rate of back propagation.")
     parser.add_argument('--enable_visdom', action='store_true',
                         help="Enable Visdom to visualize training progress")
@@ -51,12 +51,13 @@ def get_parser_for_evaluation():
     parser = argparse.ArgumentParser()
     parser.add_argument('--dataset_directory', required=True,
                         help="The location of dataset.")
-    parser.add_argument('--batch_size', type=int, default=24,
+    parser.add_argument('--batch_size', type=int, default=32,
                         help="Batch size.")
-    parser.add_argument('--data_loading_workers', type=int, default=24,
+    parser.add_argument('--data_loading_workers', type=int, default=64,
                         help="Number of workers for data loading.")
     parser.add_argument('--enable_visdom', action='store_true',
                         help="Enable Visdom to visualize training progress")
+    add_common_arguments(parser)
     return parser
 
 

data/__init__.py

@@ -1,6 +1,4 @@
-"""Defines data structure and related functions."""
-from collections import namedtuple
-
-
-MarkingPoint = namedtuple('MarkingPoint', ['x', 'y', 'direction', 'shape'])
-Slot = namedtuple('Slot', ['x1', 'y1', 'x2', 'y2'])
+"""Data related package."""
+from .data_process import get_predicted_points, match_marking_points
+from .dataset import ParkingSlotDataset
+from .struct import MarkingPoint, Slot

data/data_process.py

@@ -2,35 +2,7 @@
 import math
 import torch
 import config
-from . import MarkingPoint
-
-
-def generate_objective(marking_points_batch, device):
-    """Get regression objective and gradient for directional point detector."""
-    batch_size = len(marking_points_batch)
-    objective = torch.zeros(batch_size, config.NUM_FEATURE_MAP_CHANNEL,
-                            config.FEATURE_MAP_SIZE, config.FEATURE_MAP_SIZE,
-                            device=device)
-    gradient = torch.zeros_like(objective)
-    gradient[:, 0].fill_(1.)
-    for batch_idx, marking_points in enumerate(marking_points_batch):
-        for marking_point in marking_points:
-            col = math.floor(marking_point.x * 16)
-            row = math.floor(marking_point.y * 16)
-            # Confidence Regression
-            objective[batch_idx, 0, row, col] = 1.
-            # Makring Point Shape Regression
-            objective[batch_idx, 1, row, col] = marking_point.shape
-            # Offset Regression
-            objective[batch_idx, 2, row, col] = marking_point.x*16 - col
-            objective[batch_idx, 3, row, col] = marking_point.y*16 - row
-            # Direction Regression
-            direction = marking_point.direction
-            objective[batch_idx, 4, row, col] = math.cos(direction)
-            objective[batch_idx, 5, row, col] = math.sin(direction)
-            # Assign Gradient
-            gradient[batch_idx, 1:6, row, col].fill_(1.)
-    return objective, gradient
+from data.struct import MarkingPoint
 
 
 def non_maximum_suppression(pred_points):
@@ -55,7 +27,7 @@ def non_maximum_suppression(pred_points):
 
 
 def get_predicted_points(prediction, thresh):
-    """Get marking point from one predicted feature map."""
+    """Get marking points from one predicted feature map."""
     assert isinstance(prediction, torch.Tensor)
     predicted_points = []
     prediction = prediction.detach().cpu().numpy()

data/dataset.py

@@ -4,8 +4,8 @@ import os
 import os.path
 import cv2 as cv
 from torch.utils.data import Dataset
-from torchvision import transforms
-from . import MarkingPoint
+from torchvision.transforms import ToTensor
+from data.struct import MarkingPoint
 
 
 class ParkingSlotDataset(Dataset):
@@ -14,7 +14,7 @@ class ParkingSlotDataset(Dataset):
         super(ParkingSlotDataset, self).__init__()
         self.root = root
         self.sample_names = []
-        self.image_transform = transforms.ToTensor()
+        self.image_transform = ToTensor()
         for file in os.listdir(root):
             if file.endswith(".json"):
                 self.sample_names.append(os.path.splitext(file)[0])

data/struct.py

@@ -0,0 +1,6 @@
+"""Defines data structure."""
+from collections import namedtuple
+
+
+MarkingPoint = namedtuple('MarkingPoint', ['x', 'y', 'direction', 'shape'])
+Slot = namedtuple('Slot', ['x1', 'y1', 'x2', 'y2'])

evaluate.py

@@ -2,34 +2,37 @@
 import torch
 from torch.utils.data import DataLoader
 import config
-from data.data_process import generate_objective, get_predicted_points, match_marking_points
-from data.dataset import ParkingSlotDataset
-from model.detector import DirectionalPointDetector
-from util.log import Logger
-from util.precision_recall import calc_average_precision, calc_precision_recall
+import util
+from data import get_predicted_points, match_marking_points
+from data import ParkingSlotDataset
+from model import DirectionalPointDetector
+from train import generate_objective
 
 
 def evaluate_detector(args):
     """Evaluate directional point detector."""
     args.cuda = not args.disable_cuda and torch.cuda.is_available()
-    device = torch.device('cuda:'+str(args.gpu_id) if args.cuda else 'cpu')
+    device = torch.device('cuda:' + str(args.gpu_id) if args.cuda else 'cpu')
+    torch.set_grad_enabled(False)
 
     dp_detector = DirectionalPointDetector(
         3, args.depth_factor, config.NUM_FEATURE_MAP_CHANNEL).to(device)
     if args.detector_weights:
         dp_detector.load_state_dict(torch.load(args.detector_weights))
+    dp_detector.eval()
 
+    torch.multiprocessing.set_sharing_strategy('file_system')
     data_loader = DataLoader(ParkingSlotDataset(args.dataset_directory),
                              batch_size=args.batch_size, shuffle=True,
                              num_workers=args.data_loading_workers,
                              collate_fn=lambda x: list(zip(*x)))
-    logger = Logger()
+    logger = util.Logger(enable_visdom=args.enable_visdom)
 
     total_loss = 0
     num_evaluation = 0
     ground_truths_list = []
     predictions_list = []
-    for image, marking_points in data_loader:
+    for iter_idx, (image, marking_points) in enumerate(data_loader):
         image = torch.stack(image)
         image = image.to(device)
         ground_truths_list += list(marking_points)
@@ -42,10 +45,11 @@ def evaluate_detector(args):
 
         pred_points = [get_predicted_points(pred, 0.01) for pred in prediction]
         predictions_list += pred_points
+        logger.log(iter=iter_idx, total_loss=total_loss)
 
-    precisions, recalls = calc_precision_recall(
+    precisions, recalls = util.calc_precision_recall(
         ground_truths_list, predictions_list, match_marking_points)
-    average_precision = calc_average_precision(precisions, recalls)
+    average_precision = util.calc_average_precision(precisions, recalls)
     if args.enable_visdom:
         logger.plot_curve(precisions, recalls)
     logger.log(average_loss=total_loss / num_evaluation,

inference.py

@@ -5,8 +5,8 @@ import numpy as np
 import torch
 from torchvision.transforms import ToTensor
 import config
-from data.data_process import get_predicted_points
-from model.detector import DirectionalPointDetector
+from data import get_predicted_points
+from model import DirectionalPointDetector
 from util import Timer
 
 
@@ -92,9 +92,11 @@ def inference_detector(args):
     """Inference demo of directional point detector."""
     args.cuda = not args.disable_cuda and torch.cuda.is_available()
     device = torch.device('cuda:' + str(args.gpu_id) if args.cuda else 'cpu')
+    torch.set_grad_enabled(False)
     dp_detector = DirectionalPointDetector(
         3, args.depth_factor, config.NUM_FEATURE_MAP_CHANNEL).to(device)
     dp_detector.load_state_dict(torch.load(args.detector_weights))
+    dp_detector.eval()
     if args.mode == "image":
         detect_image(dp_detector, device, args)
     elif args.mode == "video":

model/__init__.py

@@ -0,0 +1,2 @@
+"""Network model related package."""
+from .detector import DirectionalPointDetector

train.py

@@ -1,45 +1,77 @@
 """Train directional marking point detector."""
+import math
 import random
 import torch
 from torch.utils.data import DataLoader
 import config
-from data.data_process import get_predicted_points, generate_objective
-from data.dataset import ParkingSlotDataset
-from model.detector import DirectionalPointDetector
-from util.log import Logger
-from util import tensor2im
+import data
+import util
+from model import DirectionalPointDetector
 
 
 def plot_prediction(logger, image, marking_points, prediction):
     """Plot the ground truth and prediction of a random sample in a batch."""
     rand_sample = random.randint(0, image.size(0)-1)
-    sampled_image = tensor2im(image[rand_sample])
+    sampled_image = util.tensor2im(image[rand_sample])
     logger.plot_marking_points(sampled_image, marking_points[rand_sample],
                                win_name='gt_marking_points')
-    sampled_image = tensor2im(image[rand_sample])
-    pred_points = get_predicted_points(prediction[rand_sample], 0.01)
+    sampled_image = util.tensor2im(image[rand_sample])
+    pred_points = data.get_predicted_points(prediction[rand_sample], 0.01)
     if pred_points:
         logger.plot_marking_points(sampled_image,
                                    list(list(zip(*pred_points))[1]),
                                    win_name='pred_marking_points')
 
 
+def generate_objective(marking_points_batch, device):
+    """Get regression objective and gradient for directional point detector."""
+    batch_size = len(marking_points_batch)
+    objective = torch.zeros(batch_size, config.NUM_FEATURE_MAP_CHANNEL,
+                            config.FEATURE_MAP_SIZE, config.FEATURE_MAP_SIZE,
+                            device=device)
+    gradient = torch.zeros_like(objective)
+    gradient[:, 0].fill_(1.)
+    for batch_idx, marking_points in enumerate(marking_points_batch):
+        for marking_point in marking_points:
+            col = math.floor(marking_point.x * 16)
+            row = math.floor(marking_point.y * 16)
+            # Confidence Regression
+            objective[batch_idx, 0, row, col] = 1.
+            # Makring Point Shape Regression
+            objective[batch_idx, 1, row, col] = marking_point.shape
+            # Offset Regression
+            objective[batch_idx, 2, row, col] = marking_point.x*16 - col
+            objective[batch_idx, 3, row, col] = marking_point.y*16 - row
+            # Direction Regression
+            direction = marking_point.direction
+            objective[batch_idx, 4, row, col] = math.cos(direction)
+            objective[batch_idx, 5, row, col] = math.sin(direction)
+            # Assign Gradient
+            gradient[batch_idx, 1:6, row, col].fill_(1.)
+    return objective, gradient
+
+
 def train_detector(args):
     """Train directional point detector."""
     args.cuda = not args.disable_cuda and torch.cuda.is_available()
-    device = torch.device('cuda:'+str(args.gpu_id) if args.cuda else 'cpu')
+    device = torch.device('cuda:' + str(args.gpu_id) if args.cuda else 'cpu')
+    torch.set_grad_enabled(True)
 
     dp_detector = DirectionalPointDetector(
         3, args.depth_factor, config.NUM_FEATURE_MAP_CHANNEL).to(device)
     if args.detector_weights:
+        print("Loading weights: %s" % args.detector_weights)
         dp_detector.load_state_dict(torch.load(args.detector_weights))
+    dp_detector.train()
 
     optimizer = torch.optim.Adam(dp_detector.parameters(), lr=args.lr)
     if args.optimizer_weights:
+        print("Loading weights: %s" % args.optimizer_weights)
         optimizer.load_state_dict(torch.load(args.optimizer_weights))
 
-    logger = Logger(['train_loss'] if args.enable_visdom else None)
-    data_loader = DataLoader(ParkingSlotDataset(args.dataset_directory),
+    logger = util.Logger(args.enable_visdom,
+                         ['train_loss'] if args.enable_visdom else None)
+    data_loader = DataLoader(data.ParkingSlotDataset(args.dataset_directory),
                              batch_size=args.batch_size, shuffle=True,
                              num_workers=args.data_loading_workers,
                              collate_fn=lambda x: list(zip(*x)))

util/__init__.py

@@ -1,44 +1,4 @@
-# -*- coding: utf-8 -*-
-import math
-import time
-import cv2 as cv
-import torch
-import numpy as np
-from PIL import Image
-
-
-class Timer(object):
-    """Timer."""
-
-    def __init__(self):
-        self.start_ticking = False
-        self.start = 0.
-
-    def tic(self):
-        """Start timer."""
-        self.start = time.time()
-        self.start_ticking = True
-
-    def toc(self):
-        """End timer."""
-        duration = time.time() - self.start
-        self.start_ticking = False
-        print("Time elapsed:", duration, "s.")
-
-
-def tensor2array(image_tensor, imtype=np.uint8):
-    """
-    Convert float CxHxW image tensor between [0, 1] to HxWxC numpy ndarray
-    between [0, 255]
-    """
-    assert isinstance(image_tensor, torch.Tensor)
-    image_numpy = (image_tensor.detach().cpu().numpy()) * 255.0
-    image_numpy = np.transpose(image_numpy, (1, 2, 0)).astype(imtype)
-    return image_numpy
-
-
-def tensor2im(image_tensor, imtype=np.uint8):
-    """Convert float CxHxW BGR image tensor to RGB PIL Image"""
-    image_numpy = tensor2array(image_tensor, imtype)
-    image_numpy = cv.cvtColor(image_numpy, cv.COLOR_BGR2RGB)
-    return Image.fromarray(image_numpy)
+"""Utility related package."""
+from .log import Logger
+from .precision_recall import calc_precision_recall, calc_average_precision
+from .utils import Timer, tensor2array, tensor2im

util/log.py

@@ -8,12 +8,14 @@ from PIL import ImageDraw
 class Logger():
     """Logger for training."""
 
-    def __init__(self, curve_names=None):
+    def __init__(self, enable_visdom=False, curve_names=None):
         self.curve_names = curve_names
-        if curve_names:
+        if enable_visdom:
             self.vis = Visdom()
             assert self.vis.check_connection()
             self.curve_x = np.array([0])
+        else:
+            self.curve_names = None
 
     def log(self, xval=None, win_name='loss', **kwargs):
         """Log and print the information."""

util/precision_recall.py

@@ -1,4 +1,5 @@
 """Universal procedure of calculating precision and recall."""
+import bisect
 
 
 def match_gt_with_preds(ground_truth, predictions, match_labels):
@@ -39,16 +40,20 @@ def calc_precision_recall(ground_truths_list, predictions_list, match_labels):
     """Adjust threshold to get mutiple precision recall sample."""
     true_positive_list, false_positive_list = get_confidence_list(
         ground_truths_list, predictions_list, match_labels)
+    true_positive_list = sorted(true_positive_list)
+    false_positive_list = sorted(false_positive_list)
+    thresholds = sorted(list(set(true_positive_list)))
     recalls = [0.]
     precisions = [0.]
-    thresholds = sorted(list(set(true_positive_list)))
     for thresh in reversed(thresholds):
         if thresh == 0.:
             recalls.append(1.)
             precisions.append(0.)
-        true_positives = sum(i >= thresh for i in true_positive_list)
-        false_positives = sum(i >= thresh for i in false_positive_list)
-        false_negatives = len(true_positive_list) - true_positives
+            break
+        false_negatives = bisect.bisect_left(true_positive_list, thresh)
+        true_positives = len(true_positive_list) - false_negatives
+        true_negatives = bisect.bisect_left(false_positive_list, thresh)
+        false_positives = len(false_positive_list) - true_negatives
         recalls.append(true_positives / (true_positives+false_negatives))
         precisions.append(true_positives / (true_positives + false_positives))
     return precisions, recalls

util/utils.py

@@ -0,0 +1,44 @@
+"""Utility classes and functions."""
+import math
+import time
+import cv2 as cv
+import torch
+import numpy as np
+from PIL import Image
+
+
+class Timer(object):
+    """Timer."""
+
+    def __init__(self):
+        self.start_ticking = False
+        self.start = 0.
+
+    def tic(self):
+        """Start timer."""
+        self.start = time.time()
+        self.start_ticking = True
+
+    def toc(self):
+        """End timer."""
+        duration = time.time() - self.start
+        self.start_ticking = False
+        print("Time elapsed:", duration, "s.")
+
+
+def tensor2array(image_tensor, imtype=np.uint8):
+    """
+    Convert float CxHxW image tensor between [0, 1] to HxWxC numpy ndarray
+    between [0, 255]
+    """
+    assert isinstance(image_tensor, torch.Tensor)
+    image_numpy = (image_tensor.detach().cpu().numpy()) * 255.0
+    image_numpy = np.transpose(image_numpy, (1, 2, 0)).astype(imtype)
+    return image_numpy
+
+
+def tensor2im(image_tensor, imtype=np.uint8):
+    """Convert float CxHxW BGR image tensor to RGB PIL Image"""
+    image_numpy = tensor2array(image_tensor, imtype)
+    image_numpy = cv.cvtColor(image_numpy, cv.COLOR_BGR2RGB)
+    return Image.fromarray(image_numpy)