datasets.py

import os
import numpy as np
import cv2
#import pandas as pd

import torch
from torch.utils.data.dataset import Dataset
from torchvision import transforms

from PIL import Image, ImageFilter

import utils

def get_list_from_filenames(file_path):
    # input:    relative path to .txt file with file names
    # output:   list of relative path names
    with open(file_path) as f:
        lines = f.read().splitlines()
    return lines

class BIWI_multi(Dataset):
    def __init__(self, data_dir, filename_path, transform, img_ext='.png', annot_ext='.txt', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        img = Image.open(os.path.join(self.data_dir, self.X_train[index] + '_rgb' + self.img_ext))
        img = img.convert(self.image_mode)
        pose_path = os.path.join(self.data_dir, self.y_train[index] + '_pose' + self.annot_ext)

        bbox_path = os.path.join(self.data_dir, self.y_train[index] + '_rgb' + self.annot_ext)

        # Load bounding box
        bbox = open(bbox_path, 'r')
        line = bbox.readline().split(' ')
        if len(line) < 4:
            x_min, y_min, x_max, y_max = 0, 0, img.size[0], img.size[1]
        else:
            x_min, y_min, x_max, y_max = [float(line[1]), float(line[2]), float(line[3]), float(line[4])]
        bbox.close()

        # Load pose in degrees
        pose_annot = open(pose_path, 'r')
        R = []
        for line in pose_annot:
            line = line.strip('\n').split(' ')
            l = []
            if line[0] != '':
                for nb in line:
                    if nb == '':
                        continue
                    l.append(float(nb))
                R.append(l)

        R = np.array(R)
        T = R[3,:]
        R = R[:3,:]
        pose_annot.close()

        R = np.transpose(R)

        roll = -np.arctan2(R[1][0], R[0][0]) * 180 / np.pi
        yaw = -np.arctan2(-R[2][0], np.sqrt(R[2][1] ** 2 + R[2][2] ** 2)) * 180 / np.pi
        pitch = np.arctan2(R[2][1], R[2][2]) * 180 / np.pi

        # Loosely crop face
        k = 0.35
        x_min -= 0.6 * k * abs(x_max - x_min)
        y_min -= k * abs(y_max - y_min)
        x_max += 0.6 * k * abs(x_max - x_min)
        y_max += 0.6 * k * abs(y_max - y_min)
        img = img.crop((int(x_min), int(y_min), int(x_max), int(y_max)))

       # Bin values
        bins = np.array(range(-99, 102, 1))
        binned_pose = torch.LongTensor(np.digitize([yaw, pitch, roll], bins) - 1)
        
        bins_0 = np.array(range(-99, 102, 3))
        binned_pose_0 = torch.LongTensor(np.digitize([yaw, pitch, roll], bins_0) - 1)
        
        bins_1 = np.array(range(-99, 102, 11))
        binned_pose_1 = torch.LongTensor(np.digitize([yaw, pitch, roll], bins_1) - 1)
        
        bins_2 = np.array(range(-99, 102, 33))
        binned_pose_2 = torch.LongTensor(np.digitize([yaw, pitch, roll], bins_2) - 1)
        
        bins_3 = np.array(range(-99, 102, 99))
        binned_pose_3 = torch.LongTensor(np.digitize([yaw, pitch, roll], bins_3) - 1)

        # Get target tensors
        labels = binned_pose
        labels_0 = binned_pose_0
        labels_1 = binned_pose_1
        labels_2 = binned_pose_2
        labels_3 = binned_pose_3
        
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, labels_0, labels_1, labels_2, labels_3, cont_labels, self.X_train[index]

    def __len__(self):
        # 15,667
        return self.length
    
class AFLW_multi(Dataset):
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.txt', image_mode='RGB'):
          self.data_dir = data_dir
          self.transform = transform
          self.img_ext = img_ext
          self.annot_ext = annot_ext

          filename_list = get_list_from_filenames(filename_path)

          self.X_train = filename_list
          self.y_train = filename_list
          self.image_mode = image_mode
          self.length = len(filename_list)

    def __getitem__(self, index):
        if not os.path.exists(os.path.join(self.data_dir, self.X_train[index] + self.img_ext)):
          img = Image.open(os.path.join(self.data_dir, self.X_train[index] + '.png'))
        else:
          img = Image.open(os.path.join(self.data_dir, self.X_train[index] + self.img_ext))
        img = img.convert(self.image_mode)
        txt_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # We get the pose in radians
        annot = open(txt_path, 'r')
        line = annot.readline().split('\t')
        pose = [float(line[1]), float(line[2]), float(line[3])]
        # And convert to degrees.
        yaw = pose[0] * 180 / np.pi
        pitch = pose[1] * 180 / np.pi
        roll = pose[2] * 180 / np.pi
        # Fix the roll in AFLW
        roll *= -1
        
        # Bin values
        bins = np.array(range(-99, 102, 1))
        binned_pose = torch.LongTensor(np.digitize([yaw, pitch, roll], bins) - 1)
        
        bins_0 = np.array(range(-99, 102, 3))
        binned_pose_0 = torch.LongTensor(np.digitize([yaw, pitch, roll], bins_0) - 1)
        
        bins_1 = np.array(range(-99, 102, 11))
        binned_pose_1 = torch.LongTensor(np.digitize([yaw, pitch, roll], bins_1) - 1)
        
        bins_2 = np.array(range(-99, 102, 33))
        binned_pose_2 = torch.LongTensor(np.digitize([yaw, pitch, roll], bins_2) - 1)
        
        bins_3 = np.array(range(-99, 102, 99))
        binned_pose_3 = torch.LongTensor(np.digitize([yaw, pitch, roll], bins_3) - 1)

        # Get target tensors
        labels = binned_pose
        labels_0 = binned_pose_0
        labels_1 = binned_pose_1
        labels_2 = binned_pose_2
        labels_3 = binned_pose_3
        
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, labels_0, labels_1, labels_2, labels_3, cont_labels, self.X_train[index]

    def __len__(self):
        # train: 18,863
        # test: 1,966
        return self.length
    
class Pose_300W_LP_multi(Dataset):
    # Head pose from 300W-LP dataset
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.mat', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        img = Image.open(os.path.join(self.data_dir, self.X_train[index] + self.img_ext))
        img = img.convert(self.image_mode)
        mat_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # Crop the face loosely
        pt2d = utils.get_pt2d_from_mat(mat_path)
        x_min = min(pt2d[0,:])
        y_min = min(pt2d[1,:])
        x_max = max(pt2d[0,:])
        y_max = max(pt2d[1,:])

        # k = 0.2 to 0.40
        k = np.random.random_sample() * 0.2 + 0.2
        x_min -= 0.6 * k * abs(x_max - x_min)
        y_min -= 2 * k * abs(y_max - y_min)
        x_max += 0.6 * k * abs(x_max - x_min)
        y_max += 0.6 * k * abs(y_max - y_min)
        img = img.crop((int(x_min), int(y_min), int(x_max), int(y_max)))

        # We get the pose in radians
        pose = utils.get_ypr_from_mat(mat_path)
        # And convert to degrees.
        pitch = pose[0] * 180 / np.pi
        yaw = pose[1] * 180 / np.pi
        roll = pose[2] * 180 / np.pi

        # Flip?
        rnd = np.random.random_sample()
        if rnd < 0.5:
            yaw = -yaw
            roll = -roll
            img = img.transpose(Image.FLIP_LEFT_RIGHT)

        # Blur?
        rnd = np.random.random_sample()
        if rnd < 0.05:
            img = img.filter(ImageFilter.BLUR)

        # Bin values
        bins = np.array(range(-99, 102, 1))
        binned_pose = np.digitize([yaw, pitch, roll], bins) - 1
        
        bins_0 = np.array(range(-99, 102, 3))
        binned_pose_0 = np.digitize([yaw, pitch, roll], bins_0) - 1
        
        bins_1 = np.array(range(-99, 102, 11))
        binned_pose_1 = np.digitize([yaw, pitch, roll], bins_1) - 1
        
        bins_2 = np.array(range(-99, 102, 33))
        binned_pose_2 = np.digitize([yaw, pitch, roll], bins_2) - 1
        
        bins_3 = np.array(range(-99, 102, 99))
        binned_pose_3 = np.digitize([yaw, pitch, roll], bins_3) - 1

        # Get target tensors
        labels = binned_pose
        labels_0 = binned_pose_0
        labels_1 = binned_pose_1
        labels_2 = binned_pose_2
        labels_3 = binned_pose_3
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, labels_0, labels_1, labels_2, labels_3, cont_labels, self.X_train[index]

    def __len__(self):
        # 122,450
        return self.length

class Pose_300W_LP(Dataset):
    # Head pose from 300W-LP dataset
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.mat', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        img = Image.open(os.path.join(self.data_dir, self.X_train[index] + self.img_ext))
        img = img.convert(self.image_mode)
        mat_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # Crop the face loosely
        pt2d = utils.get_pt2d_from_mat(mat_path)
        x_min = min(pt2d[0,:])
        y_min = min(pt2d[1,:])
        x_max = max(pt2d[0,:])
        y_max = max(pt2d[1,:])

        # k = 0.2 to 0.40
        k = np.random.random_sample() * 0.2 + 0.2
        x_min -= 0.6 * k * abs(x_max - x_min)
        y_min -= 2 * k * abs(y_max - y_min)
        x_max += 0.6 * k * abs(x_max - x_min)
        y_max += 0.6 * k * abs(y_max - y_min)
        img = img.crop((int(x_min), int(y_min), int(x_max), int(y_max)))

        # We get the pose in radians
        pose = utils.get_ypr_from_mat(mat_path)
        # And convert to degrees.
        pitch = pose[0] * 180 / np.pi
        yaw = pose[1] * 180 / np.pi
        roll = pose[2] * 180 / np.pi

        # Flip?
        rnd = np.random.random_sample()
        if rnd < 0.5:
            yaw = -yaw
            roll = -roll
            img = img.transpose(Image.FLIP_LEFT_RIGHT)

        # Blur?
        rnd = np.random.random_sample()
        if rnd < 0.05:
            img = img.filter(ImageFilter.BLUR)

        # Bin values
        bins = np.array(range(-99, 102, 3))
        binned_pose = np.digitize([yaw, pitch, roll], bins) - 1

        # Get target tensors
        labels = binned_pose
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, cont_labels, self.X_train[index]

    def __len__(self):
        # 122,450
        return self.length

class Pose_300W_LP_random_ds(Dataset):
    # 300W-LP dataset with random downsampling
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.mat', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        img = Image.open(os.path.join(self.data_dir, self.X_train[index] + self.img_ext))
        img = img.convert(self.image_mode)
        mat_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # Crop the face loosely
        pt2d = utils.get_pt2d_from_mat(mat_path)
        x_min = min(pt2d[0,:])
        y_min = min(pt2d[1,:])
        x_max = max(pt2d[0,:])
        y_max = max(pt2d[1,:])

        # k = 0.2 to 0.40
        k = np.random.random_sample() * 0.2 + 0.2
        x_min -= 0.6 * k * abs(x_max - x_min)
        y_min -= 2 * k * abs(y_max - y_min)
        x_max += 0.6 * k * abs(x_max - x_min)
        y_max += 0.6 * k * abs(y_max - y_min)
        img = img.crop((int(x_min), int(y_min), int(x_max), int(y_max)))

        # We get the pose in radians
        pose = utils.get_ypr_from_mat(mat_path)
        pitch = pose[0] * 180 / np.pi
        yaw = pose[1] * 180 / np.pi
        roll = pose[2] * 180 / np.pi

        ds = 1 + np.random.randint(0,4) * 5
        original_size = img.size
        img = img.resize((img.size[0] / ds, img.size[1] / ds), resample=Image.NEAREST)
        img = img.resize((original_size[0], original_size[1]), resample=Image.NEAREST)

        # Flip?
        rnd = np.random.random_sample()
        if rnd < 0.5:
            yaw = -yaw
            roll = -roll
            img = img.transpose(Image.FLIP_LEFT_RIGHT)

        # Blur?
        rnd = np.random.random_sample()
        if rnd < 0.05:
            img = img.filter(ImageFilter.BLUR)

        # Bin values
        bins = np.array(range(-99, 102, 3))
        binned_pose = np.digitize([yaw, pitch, roll], bins) - 1

        # Get target tensors
        labels = binned_pose
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, cont_labels, self.X_train[index]

    def __len__(self):
        # 122,450
        return self.length

class AFLW2000(Dataset):
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.mat', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        img = Image.open(os.path.join(self.data_dir, self.X_train[index] + self.img_ext))
        img = img.convert(self.image_mode)
        mat_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # Crop the face loosely
        pt2d = utils.get_pt2d_from_mat(mat_path)

        x_min = min(pt2d[0,:])
        y_min = min(pt2d[1,:])
        x_max = max(pt2d[0,:])
        y_max = max(pt2d[1,:])

        k = 0.20
        x_min -= 2 * k * abs(x_max - x_min)
        y_min -= 2 * k * abs(y_max - y_min)
        x_max += 2 * k * abs(x_max - x_min)
        y_max += 0.6 * k * abs(y_max - y_min)
        img = img.crop((int(x_min), int(y_min), int(x_max), int(y_max)))

        # We get the pose in radians
        pose = utils.get_ypr_from_mat(mat_path)
        # And convert to degrees.
        pitch = pose[0] * 180 / np.pi
        yaw = pose[1] * 180 / np.pi
        roll = pose[2] * 180 / np.pi
        # Bin values
        bins = np.array(range(-99, 102, 3))
        labels = torch.LongTensor(np.digitize([yaw, pitch, roll], bins) - 1)
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, cont_labels, self.X_train[index]

    def __len__(self):
        # 2,000
        return self.length

class AFLW2000_ds(Dataset):
    # AFLW2000 dataset with fixed downsampling
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.mat', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        img = Image.open(os.path.join(self.data_dir, self.X_train[index] + self.img_ext))
        img = img.convert(self.image_mode)
        mat_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # Crop the face loosely
        pt2d = utils.get_pt2d_from_mat(mat_path)
        x_min = min(pt2d[0,:])
        y_min = min(pt2d[1,:])
        x_max = max(pt2d[0,:])
        y_max = max(pt2d[1,:])

        k = 0.20
        x_min -= 2 * k * abs(x_max - x_min)
        y_min -= 2 * k * abs(y_max - y_min)
        x_max += 2 * k * abs(x_max - x_min)
        y_max += 0.6 * k * abs(y_max - y_min)
        img = img.crop((int(x_min), int(y_min), int(x_max), int(y_max)))

        ds = 3  # downsampling factor
        original_size = img.size
        img = img.resize((img.size[0] / ds, img.size[1] / ds), resample=Image.NEAREST)
        img = img.resize((original_size[0], original_size[1]), resample=Image.NEAREST)

        # We get the pose in radians
        pose = utils.get_ypr_from_mat(mat_path)
        # And convert to degrees.
        pitch = pose[0] * 180 / np.pi
        yaw = pose[1] * 180 / np.pi
        roll = pose[2] * 180 / np.pi
        # Bin values
        bins = np.array(range(-99, 102, 3))
        labels = torch.LongTensor(np.digitize([yaw, pitch, roll], bins) - 1)
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, cont_labels, self.X_train[index]

    def __len__(self):
        # 2,000
        return self.length

class AFLW_aug(Dataset):
    # AFLW dataset with flipping
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.txt', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        img = Image.open(os.path.join(self.data_dir, self.X_train[index] + self.img_ext))
        img = img.convert(self.image_mode)
        txt_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # We get the pose in radians
        annot = open(txt_path, 'r')
        line = annot.readline().split(' ')
        pose = [float(line[1]), float(line[2]), float(line[3])]
        # And convert to degrees.
        yaw = pose[0] * 180 / np.pi
        pitch = pose[1] * 180 / np.pi
        roll = pose[2] * 180 / np.pi
        # Fix the roll in AFLW
        roll *= -1

        # Augment
        # Flip?
        rnd = np.random.random_sample()
        if rnd < 0.5:
            yaw = -yaw
            roll = -roll
            img = img.transpose(Image.FLIP_LEFT_RIGHT)

        # Bin values
        bins = np.array(range(-99, 102, 3))
        labels = torch.LongTensor(np.digitize([yaw, pitch, roll], bins) - 1)
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, cont_labels, self.X_train[index]

    def __len__(self):
        # train: 18,863
        # test: 1,966
        return self.length

class AFLW(Dataset):
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.txt', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        if not os.path.exists(os.path.join(self.data_dir, self.X_train[index] + self.img_ext)):
          img = Image.open(os.path.join(self.data_dir, self.X_train[index] + '.png'))
        else:
          img = Image.open(os.path.join(self.data_dir, self.X_train[index] + self.img_ext))
        img = img.convert(self.image_mode)
        txt_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # We get the pose in radians
        annot = open(txt_path, 'r')
        line = annot.readline().split('\t')
        pose = [float(line[1]), float(line[2]), float(line[3])]
        # And convert to degrees.
        yaw = pose[0] * 180 / np.pi
        pitch = pose[1] * 180 / np.pi
        roll = pose[2] * 180 / np.pi
        # Fix the roll in AFLW
        roll *= -1
        # Bin values
        bins = np.array(range(-99, 102, 3))
        labels = torch.LongTensor(np.digitize([yaw, pitch, roll], bins) - 1)
        cont_labels = torch.FloatTensor([yaw, pitch, roll])
        print(self.y_train[index] + self.annot_ext,cont_labels,line)

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, cont_labels, self.X_train[index]

    def __len__(self):
        # train: 18,863
        # test: 1,966
        return self.length

class AFW(Dataset):
    def __init__(self, data_dir, filename_path, transform, img_ext='.jpg', annot_ext='.txt', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        txt_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)
        img_name = self.X_train[index].split('_')[0]

        img = Image.open(os.path.join(self.data_dir, img_name + self.img_ext))
        img = img.convert(self.image_mode)
        txt_path = os.path.join(self.data_dir, self.y_train[index] + self.annot_ext)

        # We get the pose in degrees
        annot = open(txt_path, 'r')
        line = annot.readline().split(' ')
        yaw, pitch, roll = [float(line[1]), float(line[2]), float(line[3])]

        # Crop the face loosely
        k = 0.32
        x1 = float(line[4])
        y1 = float(line[5])
        x2 = float(line[6])
        y2 = float(line[7])
        x1 -= 0.8 * k * abs(x2 - x1)
        y1 -= 2 * k * abs(y2 - y1)
        x2 += 0.8 * k * abs(x2 - x1)
        y2 += 1 * k * abs(y2 - y1)

        img = img.crop((int(x1), int(y1), int(x2), int(y2)))

        # Bin values
        bins = np.array(range(-99, 102, 3))
        labels = torch.LongTensor(np.digitize([yaw, pitch, roll], bins) - 1)
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, cont_labels, self.X_train[index]

    def __len__(self):
        # Around 200
        return self.length

class BIWI(Dataset):
    def __init__(self, data_dir, filename_path, transform, img_ext='.png', annot_ext='.txt', image_mode='RGB'):
        self.data_dir = data_dir
        self.transform = transform
        self.img_ext = img_ext
        self.annot_ext = annot_ext

        filename_list = get_list_from_filenames(filename_path)

        self.X_train = filename_list
        self.y_train = filename_list
        self.image_mode = image_mode
        self.length = len(filename_list)

    def __getitem__(self, index):
        img = Image.open(os.path.join(self.data_dir, self.X_train[index] + '_rgb' + self.img_ext))
        img = img.convert(self.image_mode)
        pose_path = os.path.join(self.data_dir, self.y_train[index] + '_pose' + self.annot_ext)

        #y_train_list = self.y_train[index].split('/')
        bbox_path = os.path.join(self.data_dir, self.y_train[index] + '_rgb' + self.annot_ext)

        # Load bounding box
        bbox = open(bbox_path, 'r')
        line = bbox.readline().split(' ')
        if len(line) < 4:
            x_min, y_min, x_max, y_max = 0, 0, img.size[0], img.size[1]
        else:
            x_min, y_min, x_max, y_max = [float(line[1]), float(line[2]), float(line[3]), float(line[4])]
        bbox.close()

        # Load pose in degrees
        pose_annot = open(pose_path, 'r')
        R = []
        for line in pose_annot:
            line = line.strip('\n').split(' ')
            l = []
            if line[0] != '':
                for nb in line:
                    if nb == '':
                        continue
                    l.append(float(nb))
                R.append(l)

        R = np.array(R)
        T = R[3,:]
        R = R[:3,:]
        pose_annot.close()

        R = np.transpose(R)

        roll = -np.arctan2(R[1][0], R[0][0]) * 180 / np.pi
        yaw = -np.arctan2(-R[2][0], np.sqrt(R[2][1] ** 2 + R[2][2] ** 2)) * 180 / np.pi
        pitch = np.arctan2(R[2][1], R[2][2]) * 180 / np.pi

        # Loosely crop face
        k = 0.35
        x_min -= 0.6 * k * abs(x_max - x_min)
        y_min -= k * abs(y_max - y_min)
        x_max += 0.6 * k * abs(x_max - x_min)
        y_max += 0.6 * k * abs(y_max - y_min)
        img = img.crop((int(x_min), int(y_min), int(x_max), int(y_max)))

        # Bin values
        bins = np.array(range(-99, 102, 3))
        binned_pose = np.digitize([yaw, pitch, roll], bins) - 1

        labels = torch.LongTensor(binned_pose)
        cont_labels = torch.FloatTensor([yaw, pitch, roll])

        if self.transform is not None:
            img = self.transform(img)

        return img, labels, cont_labels, self.X_train[index]

    def __len__(self):
        # 15,667
        return self.length