Upload 3 files

facelib/detection/retinaface/retinaface.py

@@ -0,0 +1,372 @@
+import cv2
+import numpy as np
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from PIL import Image
+from torchvision.models._utils import IntermediateLayerGetter as IntermediateLayerGetter
+
+from facelib.detection.align_trans import get_reference_facial_points, warp_and_crop_face
+from facelib.detection.retinaface.retinaface_net import FPN, SSH, MobileNetV1, make_bbox_head, make_class_head, make_landmark_head
+from facelib.detection.retinaface.retinaface_utils import (PriorBox, batched_decode, batched_decode_landm, decode, decode_landm,
+                                                 py_cpu_nms)
+
+from basicsr.utils.misc import get_device
+# device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+device = get_device()
+
+
+def generate_config(network_name):
+    
+    cfg_mnet = {
+        'name': 'mobilenet0.25',
+        'min_sizes': [[16, 32], [64, 128], [256, 512]],
+        'steps': [8, 16, 32],
+        'variance': [0.1, 0.2],
+        'clip': False,
+        'loc_weight': 2.0,
+        'gpu_train': True,
+        'batch_size': 32,
+        'ngpu': 1,
+        'epoch': 250,
+        'decay1': 190,
+        'decay2': 220,
+        'image_size': 640,
+        'return_layers': {
+            'stage1': 1,
+            'stage2': 2,
+            'stage3': 3
+        },
+        'in_channel': 32,
+        'out_channel': 64
+    }
+
+    cfg_re50 = {
+        'name': 'Resnet50',
+        'min_sizes': [[16, 32], [64, 128], [256, 512]],
+        'steps': [8, 16, 32],
+        'variance': [0.1, 0.2],
+        'clip': False,
+        'loc_weight': 2.0,
+        'gpu_train': True,
+        'batch_size': 24,
+        'ngpu': 4,
+        'epoch': 100,
+        'decay1': 70,
+        'decay2': 90,
+        'image_size': 840,
+        'return_layers': {
+            'layer2': 1,
+            'layer3': 2,
+            'layer4': 3
+        },
+        'in_channel': 256,
+        'out_channel': 256
+    }
+
+    if network_name == 'mobile0.25':
+        return cfg_mnet
+    elif network_name == 'resnet50':
+        return cfg_re50
+    else:
+        raise NotImplementedError(f'network_name={network_name}')
+
+
+class RetinaFace(nn.Module):
+
+    def __init__(self, network_name='resnet50', half=False, phase='test'):
+        super(RetinaFace, self).__init__()
+        self.half_inference = half
+        cfg = generate_config(network_name)
+        self.backbone = cfg['name']
+
+        self.model_name = f'retinaface_{network_name}'
+        self.cfg = cfg
+        self.phase = phase
+        self.target_size, self.max_size = 1600, 2150
+        self.resize, self.scale, self.scale1 = 1., None, None
+        self.mean_tensor = torch.tensor([[[[104.]], [[117.]], [[123.]]]]).to(device)
+        self.reference = get_reference_facial_points(default_square=True)
+        # Build network.
+        backbone = None
+        if cfg['name'] == 'mobilenet0.25':
+            backbone = MobileNetV1()
+            self.body = IntermediateLayerGetter(backbone, cfg['return_layers'])
+        elif cfg['name'] == 'Resnet50':
+            import torchvision.models as models
+            backbone = models.resnet50(pretrained=False)
+            self.body = IntermediateLayerGetter(backbone, cfg['return_layers'])
+
+        in_channels_stage2 = cfg['in_channel']
+        in_channels_list = [
+            in_channels_stage2 * 2,
+            in_channels_stage2 * 4,
+            in_channels_stage2 * 8,
+        ]
+
+        out_channels = cfg['out_channel']
+        self.fpn = FPN(in_channels_list, out_channels)
+        self.ssh1 = SSH(out_channels, out_channels)
+        self.ssh2 = SSH(out_channels, out_channels)
+        self.ssh3 = SSH(out_channels, out_channels)
+
+        self.ClassHead = make_class_head(fpn_num=3, inchannels=cfg['out_channel'])
+        self.BboxHead = make_bbox_head(fpn_num=3, inchannels=cfg['out_channel'])
+        self.LandmarkHead = make_landmark_head(fpn_num=3, inchannels=cfg['out_channel'])
+
+        self.to(device)
+        self.eval()
+        if self.half_inference:
+            self.half()
+
+    def forward(self, inputs):
+        out = self.body(inputs)
+
+        if self.backbone == 'mobilenet0.25' or self.backbone == 'Resnet50':
+            out = list(out.values())
+        # FPN
+        fpn = self.fpn(out)
+
+        # SSH
+        feature1 = self.ssh1(fpn[0])
+        feature2 = self.ssh2(fpn[1])
+        feature3 = self.ssh3(fpn[2])
+        features = [feature1, feature2, feature3]
+
+        bbox_regressions = torch.cat([self.BboxHead[i](feature) for i, feature in enumerate(features)], dim=1)
+        classifications = torch.cat([self.ClassHead[i](feature) for i, feature in enumerate(features)], dim=1)
+        tmp = [self.LandmarkHead[i](feature) for i, feature in enumerate(features)]
+        ldm_regressions = (torch.cat(tmp, dim=1))
+
+        if self.phase == 'train':
+            output = (bbox_regressions, classifications, ldm_regressions)
+        else:
+            output = (bbox_regressions, F.softmax(classifications, dim=-1), ldm_regressions)
+        return output
+
+    def __detect_faces(self, inputs):
+        # get scale
+        height, width = inputs.shape[2:]
+        self.scale = torch.tensor([width, height, width, height], dtype=torch.float32).to(device)
+        tmp = [width, height, width, height, width, height, width, height, width, height]
+        self.scale1 = torch.tensor(tmp, dtype=torch.float32).to(device)
+
+        # forawrd
+        inputs = inputs.to(device)
+        if self.half_inference:
+            inputs = inputs.half()
+        loc, conf, landmarks = self(inputs)
+
+        # get priorbox
+        priorbox = PriorBox(self.cfg, image_size=inputs.shape[2:])
+        priors = priorbox.forward().to(device)
+
+        return loc, conf, landmarks, priors
+
+    # single image detection
+    def transform(self, image, use_origin_size):
+        # convert to opencv format
+        if isinstance(image, Image.Image):
+            image = cv2.cvtColor(np.asarray(image), cv2.COLOR_RGB2BGR)
+        image = image.astype(np.float32)
+
+        # testing scale
+        im_size_min = np.min(image.shape[0:2])
+        im_size_max = np.max(image.shape[0:2])
+        resize = float(self.target_size) / float(im_size_min)
+
+        # prevent bigger axis from being more than max_size
+        if np.round(resize * im_size_max) > self.max_size:
+            resize = float(self.max_size) / float(im_size_max)
+        resize = 1 if use_origin_size else resize
+
+        # resize
+        if resize != 1:
+            image = cv2.resize(image, None, None, fx=resize, fy=resize, interpolation=cv2.INTER_LINEAR)
+
+        # convert to torch.tensor format
+        # image -= (104, 117, 123)
+        image = image.transpose(2, 0, 1)
+        image = torch.from_numpy(image).unsqueeze(0)
+
+        return image, resize
+
+    def detect_faces(
+        self,
+        image,
+        conf_threshold=0.8,
+        nms_threshold=0.4,
+        use_origin_size=True,
+    ):
+        """
+        Params:
+            imgs: BGR image
+        """
+        image, self.resize = self.transform(image, use_origin_size)
+        image = image.to(device)
+        if self.half_inference:
+            image = image.half()
+        image = image - self.mean_tensor
+
+        loc, conf, landmarks, priors = self.__detect_faces(image)
+
+        boxes = decode(loc.data.squeeze(0), priors.data, self.cfg['variance'])
+        boxes = boxes * self.scale / self.resize
+        boxes = boxes.cpu().numpy()
+
+        scores = conf.squeeze(0).data.cpu().numpy()[:, 1]
+
+        landmarks = decode_landm(landmarks.squeeze(0), priors, self.cfg['variance'])
+        landmarks = landmarks * self.scale1 / self.resize
+        landmarks = landmarks.cpu().numpy()
+
+        # ignore low scores
+        inds = np.where(scores > conf_threshold)[0]
+        boxes, landmarks, scores = boxes[inds], landmarks[inds], scores[inds]
+
+        # sort
+        order = scores.argsort()[::-1]
+        boxes, landmarks, scores = boxes[order], landmarks[order], scores[order]
+
+        # do NMS
+        bounding_boxes = np.hstack((boxes, scores[:, np.newaxis])).astype(np.float32, copy=False)
+        keep = py_cpu_nms(bounding_boxes, nms_threshold)
+        bounding_boxes, landmarks = bounding_boxes[keep, :], landmarks[keep]
+        # self.t['forward_pass'].toc()
+        # print(self.t['forward_pass'].average_time)
+        # import sys
+        # sys.stdout.flush()
+        return np.concatenate((bounding_boxes, landmarks), axis=1)
+
+    def __align_multi(self, image, boxes, landmarks, limit=None):
+
+        if len(boxes) < 1:
+            return [], []
+
+        if limit:
+            boxes = boxes[:limit]
+            landmarks = landmarks[:limit]
+
+        faces = []
+        for landmark in landmarks:
+            facial5points = [[landmark[2 * j], landmark[2 * j + 1]] for j in range(5)]
+
+            warped_face = warp_and_crop_face(np.array(image), facial5points, self.reference, crop_size=(112, 112))
+            faces.append(warped_face)
+
+        return np.concatenate((boxes, landmarks), axis=1), faces
+
+    def align_multi(self, img, conf_threshold=0.8, limit=None):
+
+        rlt = self.detect_faces(img, conf_threshold=conf_threshold)
+        boxes, landmarks = rlt[:, 0:5], rlt[:, 5:]
+
+        return self.__align_multi(img, boxes, landmarks, limit)
+
+    # batched detection
+    def batched_transform(self, frames, use_origin_size):
+        """
+        Arguments:
+            frames: a list of PIL.Image, or torch.Tensor(shape=[n, h, w, c],
+                type=np.float32, BGR format).
+            use_origin_size: whether to use origin size.
+        """
+        from_PIL = True if isinstance(frames[0], Image.Image) else False
+
+        # convert to opencv format
+        if from_PIL:
+            frames = [cv2.cvtColor(np.asarray(frame), cv2.COLOR_RGB2BGR) for frame in frames]
+            frames = np.asarray(frames, dtype=np.float32)
+
+        # testing scale
+        im_size_min = np.min(frames[0].shape[0:2])
+        im_size_max = np.max(frames[0].shape[0:2])
+        resize = float(self.target_size) / float(im_size_min)
+
+        # prevent bigger axis from being more than max_size
+        if np.round(resize * im_size_max) > self.max_size:
+            resize = float(self.max_size) / float(im_size_max)
+        resize = 1 if use_origin_size else resize
+
+        # resize
+        if resize != 1:
+            if not from_PIL:
+                frames = F.interpolate(frames, scale_factor=resize)
+            else:
+                frames = [
+                    cv2.resize(frame, None, None, fx=resize, fy=resize, interpolation=cv2.INTER_LINEAR)
+                    for frame in frames
+                ]
+
+        # convert to torch.tensor format
+        if not from_PIL:
+            frames = frames.transpose(1, 2).transpose(1, 3).contiguous()
+        else:
+            frames = frames.transpose((0, 3, 1, 2))
+            frames = torch.from_numpy(frames)
+
+        return frames, resize
+
+    def batched_detect_faces(self, frames, conf_threshold=0.8, nms_threshold=0.4, use_origin_size=True):
+        """
+        Arguments:
+            frames: a list of PIL.Image, or np.array(shape=[n, h, w, c],
+                type=np.uint8, BGR format).
+            conf_threshold: confidence threshold.
+            nms_threshold: nms threshold.
+            use_origin_size: whether to use origin size.
+        Returns:
+            final_bounding_boxes: list of np.array ([n_boxes, 5],
+                type=np.float32).
+            final_landmarks: list of np.array ([n_boxes, 10], type=np.float32).
+        """
+        # self.t['forward_pass'].tic()
+        frames, self.resize = self.batched_transform(frames, use_origin_size)
+        frames = frames.to(device)
+        frames = frames - self.mean_tensor
+
+        b_loc, b_conf, b_landmarks, priors = self.__detect_faces(frames)
+
+        final_bounding_boxes, final_landmarks = [], []
+
+        # decode
+        priors = priors.unsqueeze(0)
+        b_loc = batched_decode(b_loc, priors, self.cfg['variance']) * self.scale / self.resize
+        b_landmarks = batched_decode_landm(b_landmarks, priors, self.cfg['variance']) * self.scale1 / self.resize
+        b_conf = b_conf[:, :, 1]
+
+        # index for selection
+        b_indice = b_conf > conf_threshold
+
+        # concat
+        b_loc_and_conf = torch.cat((b_loc, b_conf.unsqueeze(-1)), dim=2).float()
+
+        for pred, landm, inds in zip(b_loc_and_conf, b_landmarks, b_indice):
+
+            # ignore low scores
+            pred, landm = pred[inds, :], landm[inds, :]
+            if pred.shape[0] == 0:
+                final_bounding_boxes.append(np.array([], dtype=np.float32))
+                final_landmarks.append(np.array([], dtype=np.float32))
+                continue
+
+            # sort
+            # order = score.argsort(descending=True)
+            # box, landm, score = box[order], landm[order], score[order]
+
+            # to CPU
+            bounding_boxes, landm = pred.cpu().numpy(), landm.cpu().numpy()
+
+            # NMS
+            keep = py_cpu_nms(bounding_boxes, nms_threshold)
+            bounding_boxes, landmarks = bounding_boxes[keep, :], landm[keep]
+
+            # append
+            final_bounding_boxes.append(bounding_boxes)
+            final_landmarks.append(landmarks)
+        # self.t['forward_pass'].toc(average=True)
+        # self.batch_time += self.t['forward_pass'].diff
+        # self.total_frame += len(frames)
+        # print(self.batch_time / self.total_frame)
+
+        return final_bounding_boxes, final_landmarks

facelib/detection/retinaface/retinaface_net.py

@@ -0,0 +1,196 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+def conv_bn(inp, oup, stride=1, leaky=0):
+    return nn.Sequential(
+        nn.Conv2d(inp, oup, 3, stride, 1, bias=False), nn.BatchNorm2d(oup),
+        nn.LeakyReLU(negative_slope=leaky, inplace=True))
+
+
+def conv_bn_no_relu(inp, oup, stride):
+    return nn.Sequential(
+        nn.Conv2d(inp, oup, 3, stride, 1, bias=False),
+        nn.BatchNorm2d(oup),
+    )
+
+
+def conv_bn1X1(inp, oup, stride, leaky=0):
+    return nn.Sequential(
+        nn.Conv2d(inp, oup, 1, stride, padding=0, bias=False), nn.BatchNorm2d(oup),
+        nn.LeakyReLU(negative_slope=leaky, inplace=True))
+
+
+def conv_dw(inp, oup, stride, leaky=0.1):
+    return nn.Sequential(
+        nn.Conv2d(inp, inp, 3, stride, 1, groups=inp, bias=False),
+        nn.BatchNorm2d(inp),
+        nn.LeakyReLU(negative_slope=leaky, inplace=True),
+        nn.Conv2d(inp, oup, 1, 1, 0, bias=False),
+        nn.BatchNorm2d(oup),
+        nn.LeakyReLU(negative_slope=leaky, inplace=True),
+    )
+
+
+class SSH(nn.Module):
+
+    def __init__(self, in_channel, out_channel):
+        super(SSH, self).__init__()
+        assert out_channel % 4 == 0
+        leaky = 0
+        if (out_channel <= 64):
+            leaky = 0.1
+        self.conv3X3 = conv_bn_no_relu(in_channel, out_channel // 2, stride=1)
+
+        self.conv5X5_1 = conv_bn(in_channel, out_channel // 4, stride=1, leaky=leaky)
+        self.conv5X5_2 = conv_bn_no_relu(out_channel // 4, out_channel // 4, stride=1)
+
+        self.conv7X7_2 = conv_bn(out_channel // 4, out_channel // 4, stride=1, leaky=leaky)
+        self.conv7x7_3 = conv_bn_no_relu(out_channel // 4, out_channel // 4, stride=1)
+
+    def forward(self, input):
+        conv3X3 = self.conv3X3(input)
+
+        conv5X5_1 = self.conv5X5_1(input)
+        conv5X5 = self.conv5X5_2(conv5X5_1)
+
+        conv7X7_2 = self.conv7X7_2(conv5X5_1)
+        conv7X7 = self.conv7x7_3(conv7X7_2)
+
+        out = torch.cat([conv3X3, conv5X5, conv7X7], dim=1)
+        out = F.relu(out)
+        return out
+
+
+class FPN(nn.Module):
+
+    def __init__(self, in_channels_list, out_channels):
+        super(FPN, self).__init__()
+        leaky = 0
+        if (out_channels <= 64):
+            leaky = 0.1
+        self.output1 = conv_bn1X1(in_channels_list[0], out_channels, stride=1, leaky=leaky)
+        self.output2 = conv_bn1X1(in_channels_list[1], out_channels, stride=1, leaky=leaky)
+        self.output3 = conv_bn1X1(in_channels_list[2], out_channels, stride=1, leaky=leaky)
+
+        self.merge1 = conv_bn(out_channels, out_channels, leaky=leaky)
+        self.merge2 = conv_bn(out_channels, out_channels, leaky=leaky)
+
+    def forward(self, input):
+        # names = list(input.keys())
+        # input = list(input.values())
+
+        output1 = self.output1(input[0])
+        output2 = self.output2(input[1])
+        output3 = self.output3(input[2])
+
+        up3 = F.interpolate(output3, size=[output2.size(2), output2.size(3)], mode='nearest')
+        output2 = output2 + up3
+        output2 = self.merge2(output2)
+
+        up2 = F.interpolate(output2, size=[output1.size(2), output1.size(3)], mode='nearest')
+        output1 = output1 + up2
+        output1 = self.merge1(output1)
+
+        out = [output1, output2, output3]
+        return out
+
+
+class MobileNetV1(nn.Module):
+
+    def __init__(self):
+        super(MobileNetV1, self).__init__()
+        self.stage1 = nn.Sequential(
+            conv_bn(3, 8, 2, leaky=0.1),  # 3
+            conv_dw(8, 16, 1),  # 7
+            conv_dw(16, 32, 2),  # 11
+            conv_dw(32, 32, 1),  # 19
+            conv_dw(32, 64, 2),  # 27
+            conv_dw(64, 64, 1),  # 43
+        )
+        self.stage2 = nn.Sequential(
+            conv_dw(64, 128, 2),  # 43 + 16 = 59
+            conv_dw(128, 128, 1),  # 59 + 32 = 91
+            conv_dw(128, 128, 1),  # 91 + 32 = 123
+            conv_dw(128, 128, 1),  # 123 + 32 = 155
+            conv_dw(128, 128, 1),  # 155 + 32 = 187
+            conv_dw(128, 128, 1),  # 187 + 32 = 219
+        )
+        self.stage3 = nn.Sequential(
+            conv_dw(128, 256, 2),  # 219 +3 2 = 241
+            conv_dw(256, 256, 1),  # 241 + 64 = 301
+        )
+        self.avg = nn.AdaptiveAvgPool2d((1, 1))
+        self.fc = nn.Linear(256, 1000)
+
+    def forward(self, x):
+        x = self.stage1(x)
+        x = self.stage2(x)
+        x = self.stage3(x)
+        x = self.avg(x)
+        # x = self.model(x)
+        x = x.view(-1, 256)
+        x = self.fc(x)
+        return x
+
+
+class ClassHead(nn.Module):
+
+    def __init__(self, inchannels=512, num_anchors=3):
+        super(ClassHead, self).__init__()
+        self.num_anchors = num_anchors
+        self.conv1x1 = nn.Conv2d(inchannels, self.num_anchors * 2, kernel_size=(1, 1), stride=1, padding=0)
+
+    def forward(self, x):
+        out = self.conv1x1(x)
+        out = out.permute(0, 2, 3, 1).contiguous()
+
+        return out.view(out.shape[0], -1, 2)
+
+
+class BboxHead(nn.Module):
+
+    def __init__(self, inchannels=512, num_anchors=3):
+        super(BboxHead, self).__init__()
+        self.conv1x1 = nn.Conv2d(inchannels, num_anchors * 4, kernel_size=(1, 1), stride=1, padding=0)
+
+    def forward(self, x):
+        out = self.conv1x1(x)
+        out = out.permute(0, 2, 3, 1).contiguous()
+
+        return out.view(out.shape[0], -1, 4)
+
+
+class LandmarkHead(nn.Module):
+
+    def __init__(self, inchannels=512, num_anchors=3):
+        super(LandmarkHead, self).__init__()
+        self.conv1x1 = nn.Conv2d(inchannels, num_anchors * 10, kernel_size=(1, 1), stride=1, padding=0)
+
+    def forward(self, x):
+        out = self.conv1x1(x)
+        out = out.permute(0, 2, 3, 1).contiguous()
+
+        return out.view(out.shape[0], -1, 10)
+
+
+def make_class_head(fpn_num=3, inchannels=64, anchor_num=2):
+    classhead = nn.ModuleList()
+    for i in range(fpn_num):
+        classhead.append(ClassHead(inchannels, anchor_num))
+    return classhead
+
+
+def make_bbox_head(fpn_num=3, inchannels=64, anchor_num=2):
+    bboxhead = nn.ModuleList()
+    for i in range(fpn_num):
+        bboxhead.append(BboxHead(inchannels, anchor_num))
+    return bboxhead
+
+
+def make_landmark_head(fpn_num=3, inchannels=64, anchor_num=2):
+    landmarkhead = nn.ModuleList()
+    for i in range(fpn_num):
+        landmarkhead.append(LandmarkHead(inchannels, anchor_num))
+    return landmarkhead

facelib/detection/retinaface/retinaface_utils.py

@@ -0,0 +1,421 @@
+import numpy as np
+import torch
+import torchvision
+from itertools import product as product
+from math import ceil
+
+
+class PriorBox(object):
+
+    def __init__(self, cfg, image_size=None, phase='train'):
+        super(PriorBox, self).__init__()
+        self.min_sizes = cfg['min_sizes']
+        self.steps = cfg['steps']
+        self.clip = cfg['clip']
+        self.image_size = image_size
+        self.feature_maps = [[ceil(self.image_size[0] / step), ceil(self.image_size[1] / step)] for step in self.steps]
+        self.name = 's'
+
+    def forward(self):
+        anchors = []
+        for k, f in enumerate(self.feature_maps):
+            min_sizes = self.min_sizes[k]
+            for i, j in product(range(f[0]), range(f[1])):
+                for min_size in min_sizes:
+                    s_kx = min_size / self.image_size[1]
+                    s_ky = min_size / self.image_size[0]
+                    dense_cx = [x * self.steps[k] / self.image_size[1] for x in [j + 0.5]]
+                    dense_cy = [y * self.steps[k] / self.image_size[0] for y in [i + 0.5]]
+                    for cy, cx in product(dense_cy, dense_cx):
+                        anchors += [cx, cy, s_kx, s_ky]
+
+        # back to torch land
+        output = torch.Tensor(anchors).view(-1, 4)
+        if self.clip:
+            output.clamp_(max=1, min=0)
+        return output
+
+
+def py_cpu_nms(dets, thresh):
+    """Pure Python NMS baseline."""
+    keep = torchvision.ops.nms(
+        boxes=torch.Tensor(dets[:, :4]),
+        scores=torch.Tensor(dets[:, 4]),
+        iou_threshold=thresh,
+    )
+
+    return list(keep)
+
+
+def point_form(boxes):
+    """ Convert prior_boxes to (xmin, ymin, xmax, ymax)
+    representation for comparison to point form ground truth data.
+    Args:
+        boxes: (tensor) center-size default boxes from priorbox layers.
+    Return:
+        boxes: (tensor) Converted xmin, ymin, xmax, ymax form of boxes.
+    """
+    return torch.cat(
+        (
+            boxes[:, :2] - boxes[:, 2:] / 2,  # xmin, ymin
+            boxes[:, :2] + boxes[:, 2:] / 2),
+        1)  # xmax, ymax
+
+
+def center_size(boxes):
+    """ Convert prior_boxes to (cx, cy, w, h)
+    representation for comparison to center-size form ground truth data.
+    Args:
+        boxes: (tensor) point_form boxes
+    Return:
+        boxes: (tensor) Converted xmin, ymin, xmax, ymax form of boxes.
+    """
+    return torch.cat(
+        (boxes[:, 2:] + boxes[:, :2]) / 2,  # cx, cy
+        boxes[:, 2:] - boxes[:, :2],
+        1)  # w, h
+
+
+def intersect(box_a, box_b):
+    """ We resize both tensors to [A,B,2] without new malloc:
+    [A,2] -> [A,1,2] -> [A,B,2]
+    [B,2] -> [1,B,2] -> [A,B,2]
+    Then we compute the area of intersect between box_a and box_b.
+    Args:
+      box_a: (tensor) bounding boxes, Shape: [A,4].
+      box_b: (tensor) bounding boxes, Shape: [B,4].
+    Return:
+      (tensor) intersection area, Shape: [A,B].
+    """
+    A = box_a.size(0)
+    B = box_b.size(0)
+    max_xy = torch.min(box_a[:, 2:].unsqueeze(1).expand(A, B, 2), box_b[:, 2:].unsqueeze(0).expand(A, B, 2))
+    min_xy = torch.max(box_a[:, :2].unsqueeze(1).expand(A, B, 2), box_b[:, :2].unsqueeze(0).expand(A, B, 2))
+    inter = torch.clamp((max_xy - min_xy), min=0)
+    return inter[:, :, 0] * inter[:, :, 1]
+
+
+def jaccard(box_a, box_b):
+    """Compute the jaccard overlap of two sets of boxes.  The jaccard overlap
+    is simply the intersection over union of two boxes.  Here we operate on
+    ground truth boxes and default boxes.
+    E.g.:
+        A ∩ B / A ∪ B = A ∩ B / (area(A) + area(B) - A ∩ B)
+    Args:
+        box_a: (tensor) Ground truth bounding boxes, Shape: [num_objects,4]
+        box_b: (tensor) Prior boxes from priorbox layers, Shape: [num_priors,4]
+    Return:
+        jaccard overlap: (tensor) Shape: [box_a.size(0), box_b.size(0)]
+    """
+    inter = intersect(box_a, box_b)
+    area_a = ((box_a[:, 2] - box_a[:, 0]) * (box_a[:, 3] - box_a[:, 1])).unsqueeze(1).expand_as(inter)  # [A,B]
+    area_b = ((box_b[:, 2] - box_b[:, 0]) * (box_b[:, 3] - box_b[:, 1])).unsqueeze(0).expand_as(inter)  # [A,B]
+    union = area_a + area_b - inter
+    return inter / union  # [A,B]
+
+
+def matrix_iou(a, b):
+    """
+    return iou of a and b, numpy version for data augenmentation
+    """
+    lt = np.maximum(a[:, np.newaxis, :2], b[:, :2])
+    rb = np.minimum(a[:, np.newaxis, 2:], b[:, 2:])
+
+    area_i = np.prod(rb - lt, axis=2) * (lt < rb).all(axis=2)
+    area_a = np.prod(a[:, 2:] - a[:, :2], axis=1)
+    area_b = np.prod(b[:, 2:] - b[:, :2], axis=1)
+    return area_i / (area_a[:, np.newaxis] + area_b - area_i)
+
+
+def matrix_iof(a, b):
+    """
+    return iof of a and b, numpy version for data augenmentation
+    """
+    lt = np.maximum(a[:, np.newaxis, :2], b[:, :2])
+    rb = np.minimum(a[:, np.newaxis, 2:], b[:, 2:])
+
+    area_i = np.prod(rb - lt, axis=2) * (lt < rb).all(axis=2)
+    area_a = np.prod(a[:, 2:] - a[:, :2], axis=1)
+    return area_i / np.maximum(area_a[:, np.newaxis], 1)
+
+
+def match(threshold, truths, priors, variances, labels, landms, loc_t, conf_t, landm_t, idx):
+    """Match each prior box with the ground truth box of the highest jaccard
+    overlap, encode the bounding boxes, then return the matched indices
+    corresponding to both confidence and location preds.
+    Args:
+        threshold: (float) The overlap threshold used when matching boxes.
+        truths: (tensor) Ground truth boxes, Shape: [num_obj, 4].
+        priors: (tensor) Prior boxes from priorbox layers, Shape: [n_priors,4].
+        variances: (tensor) Variances corresponding to each prior coord,
+            Shape: [num_priors, 4].
+        labels: (tensor) All the class labels for the image, Shape: [num_obj].
+        landms: (tensor) Ground truth landms, Shape [num_obj, 10].
+        loc_t: (tensor) Tensor to be filled w/ encoded location targets.
+        conf_t: (tensor) Tensor to be filled w/ matched indices for conf preds.
+        landm_t: (tensor) Tensor to be filled w/ encoded landm targets.
+        idx: (int) current batch index
+    Return:
+        The matched indices corresponding to 1)location 2)confidence
+        3)landm preds.
+    """
+    # jaccard index
+    overlaps = jaccard(truths, point_form(priors))
+    # (Bipartite Matching)
+    # [1,num_objects] best prior for each ground truth
+    best_prior_overlap, best_prior_idx = overlaps.max(1, keepdim=True)
+
+    # ignore hard gt
+    valid_gt_idx = best_prior_overlap[:, 0] >= 0.2
+    best_prior_idx_filter = best_prior_idx[valid_gt_idx, :]
+    if best_prior_idx_filter.shape[0] <= 0:
+        loc_t[idx] = 0
+        conf_t[idx] = 0
+        return
+
+    # [1,num_priors] best ground truth for each prior
+    best_truth_overlap, best_truth_idx = overlaps.max(0, keepdim=True)
+    best_truth_idx.squeeze_(0)
+    best_truth_overlap.squeeze_(0)
+    best_prior_idx.squeeze_(1)
+    best_prior_idx_filter.squeeze_(1)
+    best_prior_overlap.squeeze_(1)
+    best_truth_overlap.index_fill_(0, best_prior_idx_filter, 2)  # ensure best prior
+    # TODO refactor: index  best_prior_idx with long tensor
+    # ensure every gt matches with its prior of max overlap
+    for j in range(best_prior_idx.size(0)):  # 判别此anchor是预测哪一个boxes
+        best_truth_idx[best_prior_idx[j]] = j
+    matches = truths[best_truth_idx]  # Shape: [num_priors,4] 此处为每一个anchor对应的bbox取出来
+    conf = labels[best_truth_idx]  # Shape: [num_priors]      此处为每一个anchor对应的label取出来
+    conf[best_truth_overlap < threshold] = 0  # label as background   overlap<0.35的全部作为负样本
+    loc = encode(matches, priors, variances)
+
+    matches_landm = landms[best_truth_idx]
+    landm = encode_landm(matches_landm, priors, variances)
+    loc_t[idx] = loc  # [num_priors,4] encoded offsets to learn
+    conf_t[idx] = conf  # [num_priors] top class label for each prior
+    landm_t[idx] = landm
+
+
+def encode(matched, priors, variances):
+    """Encode the variances from the priorbox layers into the ground truth boxes
+    we have matched (based on jaccard overlap) with the prior boxes.
+    Args:
+        matched: (tensor) Coords of ground truth for each prior in point-form
+            Shape: [num_priors, 4].
+        priors: (tensor) Prior boxes in center-offset form
+            Shape: [num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        encoded boxes (tensor), Shape: [num_priors, 4]
+    """
+
+    # dist b/t match center and prior's center
+    g_cxcy = (matched[:, :2] + matched[:, 2:]) / 2 - priors[:, :2]
+    # encode variance
+    g_cxcy /= (variances[0] * priors[:, 2:])
+    # match wh / prior wh
+    g_wh = (matched[:, 2:] - matched[:, :2]) / priors[:, 2:]
+    g_wh = torch.log(g_wh) / variances[1]
+    # return target for smooth_l1_loss
+    return torch.cat([g_cxcy, g_wh], 1)  # [num_priors,4]
+
+
+def encode_landm(matched, priors, variances):
+    """Encode the variances from the priorbox layers into the ground truth boxes
+    we have matched (based on jaccard overlap) with the prior boxes.
+    Args:
+        matched: (tensor) Coords of ground truth for each prior in point-form
+            Shape: [num_priors, 10].
+        priors: (tensor) Prior boxes in center-offset form
+            Shape: [num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        encoded landm (tensor), Shape: [num_priors, 10]
+    """
+
+    # dist b/t match center and prior's center
+    matched = torch.reshape(matched, (matched.size(0), 5, 2))
+    priors_cx = priors[:, 0].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2)
+    priors_cy = priors[:, 1].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2)
+    priors_w = priors[:, 2].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2)
+    priors_h = priors[:, 3].unsqueeze(1).expand(matched.size(0), 5).unsqueeze(2)
+    priors = torch.cat([priors_cx, priors_cy, priors_w, priors_h], dim=2)
+    g_cxcy = matched[:, :, :2] - priors[:, :, :2]
+    # encode variance
+    g_cxcy /= (variances[0] * priors[:, :, 2:])
+    # g_cxcy /= priors[:, :, 2:]
+    g_cxcy = g_cxcy.reshape(g_cxcy.size(0), -1)
+    # return target for smooth_l1_loss
+    return g_cxcy
+
+
+# Adapted from https://github.com/Hakuyume/chainer-ssd
+def decode(loc, priors, variances):
+    """Decode locations from predictions using priors to undo
+    the encoding we did for offset regression at train time.
+    Args:
+        loc (tensor): location predictions for loc layers,
+            Shape: [num_priors,4]
+        priors (tensor): Prior boxes in center-offset form.
+            Shape: [num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        decoded bounding box predictions
+    """
+
+    boxes = torch.cat((priors[:, :2] + loc[:, :2] * variances[0] * priors[:, 2:],
+                       priors[:, 2:] * torch.exp(loc[:, 2:] * variances[1])), 1)
+    boxes[:, :2] -= boxes[:, 2:] / 2
+    boxes[:, 2:] += boxes[:, :2]
+    return boxes
+
+
+def decode_landm(pre, priors, variances):
+    """Decode landm from predictions using priors to undo
+    the encoding we did for offset regression at train time.
+    Args:
+        pre (tensor): landm predictions for loc layers,
+            Shape: [num_priors,10]
+        priors (tensor): Prior boxes in center-offset form.
+            Shape: [num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        decoded landm predictions
+    """
+    tmp = (
+        priors[:, :2] + pre[:, :2] * variances[0] * priors[:, 2:],
+        priors[:, :2] + pre[:, 2:4] * variances[0] * priors[:, 2:],
+        priors[:, :2] + pre[:, 4:6] * variances[0] * priors[:, 2:],
+        priors[:, :2] + pre[:, 6:8] * variances[0] * priors[:, 2:],
+        priors[:, :2] + pre[:, 8:10] * variances[0] * priors[:, 2:],
+    )
+    landms = torch.cat(tmp, dim=1)
+    return landms
+
+
+def batched_decode(b_loc, priors, variances):
+    """Decode locations from predictions using priors to undo
+    the encoding we did for offset regression at train time.
+    Args:
+        b_loc (tensor): location predictions for loc layers,
+            Shape: [num_batches,num_priors,4]
+        priors (tensor): Prior boxes in center-offset form.
+            Shape: [1,num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        decoded bounding box predictions
+    """
+    boxes = (
+        priors[:, :, :2] + b_loc[:, :, :2] * variances[0] * priors[:, :, 2:],
+        priors[:, :, 2:] * torch.exp(b_loc[:, :, 2:] * variances[1]),
+    )
+    boxes = torch.cat(boxes, dim=2)
+
+    boxes[:, :, :2] -= boxes[:, :, 2:] / 2
+    boxes[:, :, 2:] += boxes[:, :, :2]
+    return boxes
+
+
+def batched_decode_landm(pre, priors, variances):
+    """Decode landm from predictions using priors to undo
+    the encoding we did for offset regression at train time.
+    Args:
+        pre (tensor): landm predictions for loc layers,
+            Shape: [num_batches,num_priors,10]
+        priors (tensor): Prior boxes in center-offset form.
+            Shape: [1,num_priors,4].
+        variances: (list[float]) Variances of priorboxes
+    Return:
+        decoded landm predictions
+    """
+    landms = (
+        priors[:, :, :2] + pre[:, :, :2] * variances[0] * priors[:, :, 2:],
+        priors[:, :, :2] + pre[:, :, 2:4] * variances[0] * priors[:, :, 2:],
+        priors[:, :, :2] + pre[:, :, 4:6] * variances[0] * priors[:, :, 2:],
+        priors[:, :, :2] + pre[:, :, 6:8] * variances[0] * priors[:, :, 2:],
+        priors[:, :, :2] + pre[:, :, 8:10] * variances[0] * priors[:, :, 2:],
+    )
+    landms = torch.cat(landms, dim=2)
+    return landms
+
+
+def log_sum_exp(x):
+    """Utility function for computing log_sum_exp while determining
+    This will be used to determine unaveraged confidence loss across
+    all examples in a batch.
+    Args:
+        x (Variable(tensor)): conf_preds from conf layers
+    """
+    x_max = x.data.max()
+    return torch.log(torch.sum(torch.exp(x - x_max), 1, keepdim=True)) + x_max
+
+
+# Original author: Francisco Massa:
+# https://github.com/fmassa/object-detection.torch
+# Ported to PyTorch by Max deGroot (02/01/2017)
+def nms(boxes, scores, overlap=0.5, top_k=200):
+    """Apply non-maximum suppression at test time to avoid detecting too many
+    overlapping bounding boxes for a given object.
+    Args:
+        boxes: (tensor) The location preds for the img, Shape: [num_priors,4].
+        scores: (tensor) The class predscores for the img, Shape:[num_priors].
+        overlap: (float) The overlap thresh for suppressing unnecessary boxes.
+        top_k: (int) The Maximum number of box preds to consider.
+    Return:
+        The indices of the kept boxes with respect to num_priors.
+    """
+
+    keep = torch.Tensor(scores.size(0)).fill_(0).long()
+    if boxes.numel() == 0:
+        return keep
+    x1 = boxes[:, 0]
+    y1 = boxes[:, 1]
+    x2 = boxes[:, 2]
+    y2 = boxes[:, 3]
+    area = torch.mul(x2 - x1, y2 - y1)
+    v, idx = scores.sort(0)  # sort in ascending order
+    # I = I[v >= 0.01]
+    idx = idx[-top_k:]  # indices of the top-k largest vals
+    xx1 = boxes.new()
+    yy1 = boxes.new()
+    xx2 = boxes.new()
+    yy2 = boxes.new()
+    w = boxes.new()
+    h = boxes.new()
+
+    # keep = torch.Tensor()
+    count = 0
+    while idx.numel() > 0:
+        i = idx[-1]  # index of current largest val
+        # keep.append(i)
+        keep[count] = i
+        count += 1
+        if idx.size(0) == 1:
+            break
+        idx = idx[:-1]  # remove kept element from view
+        # load bboxes of next highest vals
+        torch.index_select(x1, 0, idx, out=xx1)
+        torch.index_select(y1, 0, idx, out=yy1)
+        torch.index_select(x2, 0, idx, out=xx2)
+        torch.index_select(y2, 0, idx, out=yy2)
+        # store element-wise max with next highest score
+        xx1 = torch.clamp(xx1, min=x1[i])
+        yy1 = torch.clamp(yy1, min=y1[i])
+        xx2 = torch.clamp(xx2, max=x2[i])
+        yy2 = torch.clamp(yy2, max=y2[i])
+        w.resize_as_(xx2)
+        h.resize_as_(yy2)
+        w = xx2 - xx1
+        h = yy2 - yy1
+        # check sizes of xx1 and xx2.. after each iteration
+        w = torch.clamp(w, min=0.0)
+        h = torch.clamp(h, min=0.0)
+        inter = w * h
+        # IoU = i / (area(a) + area(b) - i)
+        rem_areas = torch.index_select(area, 0, idx)  # load remaining areas)
+        union = (rem_areas - inter) + area[i]
+        IoU = inter / union  # store result in iou
+        # keep only elements with an IoU <= overlap
+        idx = idx[IoU.le(overlap)]
+    return keep, count