Normalizing to unit length along the specified dimension. Args: x: pytorch Variable Returns: x: pytorch Variable, same shape as in

L2-normalize columns of X

Parameters ---------- block (nn.Module): Bottleneck class. - For SENet154: SEBottleneck - For SE-ResN

Get the image names in a dir. Optional to return numpy array, paths.

Get the person id or cam from an image name.

1x1 convolution

Shuffle the feature map according to height dimension.

Args: path: a dir, or result of `osp.dirname(osp.abspath(file_path))` Note: `osp.exists('')` returns `False`, while `osp.exists('.')` ret

Args: index (int): Index Returns: tuple: (image, target) where target is class_index of the target class.

Returns cosine similarity between x1 and x2, computed along dim.

Compute weighted text embeddings :param image_embeddings: Tensor with dtype torch.float32, [n_img, n_region, d] :param text_embeddings: T

3x3 convolution with padding

Extract features from images in a target directory :param img_dir: the input image directory :param model_path: path to ReID model :p

query: (batch, queryL, d) context: (batch, sourceL, d)

Split feature map according to height dimension. :param x: Tensor with dtype torch.float32, [batchsize, num_channels, height, width]

Normalize a N-D numpy array along the specified axis. :param nparry: the array of vectors to be normalized :param order: order of the nor

Get the person id or cam from an image name.

Get the person id or cam from an image name.

Recover the feature map to the original order.

Create dir and save file.

Cross-Modal Projection Classfication loss(CMPC) :param image_embeddings: Tensor with dtype torch.float32 :param text_embeddin

Cross-Modal Projection Matching Loss(CMPM) :param image_embeddings: Tensor with dtype torch.float32 :param text_embeddings: T

3x3 convolution with padding

Draw the rank list :param query_name: path of the query image :param gallery_name: path of the gallery image "param output_size: the

Evaluation with market1501 metric Key: for each query identity, its gallery images from the same camera view are discarded.

consider the confidence g(x) for each fragment as equal sigma_{j} (xi - xj) = sigma_{j} xi - sigma_{j} xj attn: (batch, queryL, sourceL)

consider the confidence g(x) for each fragment as the sqrt of their similarity probability to the query fragment sigma_{j} (xi - xj)gj =