2D convolution with quantization (float32-->int8)

Convert coordinates for axis-aligned 2D boxes between two coordinate formats. Creates a copy of `tensor`, i.e. does not operate in place. C

A TensorFlow layer to create an output tensor containing anchor box coordinates and variances based on the input tensor and the pas

depthwise convolution and pointwise convolution

Computes the intersection-over-union similarity (also known as Jaccard similarity) of two sets of axis-aligned 2D rectangular boxes. L

Returns: The number of images in the dataset.

All default argument values follow the Caffe implementation. Arguments: confidence_thresh (float, optional): A float in [0,1), the

Takes a list or Numpy array of decoded predictions and applies a given list of transforms to them. The list of inverse transforms would usual

Generates batches of samples and (optionally) corresponding labels indefinitely. Can shuffle the samples consistently after each co

The same greedy non-maximum suppression algorithm as above, but slightly modified for use as an internal function for per-class NMS in `dec

The same greedy non-maximum suppression algorithm as above, but slightly modified for use as an internal function in `decode_detections_fas

The same greedy non-maximum suppression algorithm as above, but slightly modified for use as an internal function for per-class NMS in `dec

Computes the average precision for each class. Can compute the Pascal-VOC-style average precision in both the pre-2010 (k-point sam

Computes the mean average precision over all classes. Note that `compute_average_precisions()` must be called before calling this m

Computes the precisions and recalls for all classes. Note that `match_predictions()` must be called before calling this method.

Convert model prediction output back to a format that contains only the positive box predictions (i.e. the same format that `SSDInputEncode

Computes an array of the spatial positions and sizes of the anchor boxes for one predictor layer of size `feature_map_size == [feat

Produces an encoding template for the ground truth label tensor for a given batch. Note that all tensor creation, reshaping and con

Builds dictionaries that map between MS COCO category IDs, transformed category IDs, and category names. The original MS COCO category IDs

Counts the number of ground truth boxes for each class across the dataset. Arguments: ignore_neutral_boxes (bool, opti

The same as 'intersection_area()' but for internal use, i.e. without all the safety checks.

resize image with unchanged aspect ratio using padding

Loads an HDF5 dataset that is in the format that the `create_hdf5_dataset()` method produces. Arguments: verb

Returns a bipartite matching according to the given weight matrix. The algorithm works as follows: Let the first axis of `weight_matri

Matches all elements along the second axis of `weight_matrix` to their best matches along the first axis subject to the constraint that the

Matches predictions to ground truth boxes. Note that `predict_on_dataset()` must be called before calling this method. Arg

This is an JSON parser for the MS COCO datasets. It might be applicable to other datasets with minor changes to the code, but in it

This is an XML parser for the Pascal VOC datasets. It might be applicable to other datasets with minor changes to the code, but in

Runs detection predictions over the whole dataset given a model and saves them in a JSON file in the MS COCO detection results format.

Runs predictions for the given model over the entire dataset given by `data_generator`. Arguments: img_height (int): T

Build a Keras model with SSD300 architecture, see references. The base network is a reduced atrous VGG-16, extended by the SSD architecture

Build a TensorFlow model with SSD512 architecture, see references. The base network is a reduced atrous VGG-16, extended by the SSD