Multi-Resolution Cascades for Multiclass Object Detection

Authors: Mohammad Saberian, Nuno Vasconcelos

NeurIPS 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on multiclass object detection show improved performance over previous solutions.
Researcher Affiliation Collaboration Mohammad Saberian Yahoo! Labs saberian@yahoo-inc.com Nuno Vasconcelos Statistical Visual Computing Laboratory University of California, San Diego nuno@ucsd.edu
Pseudocode No The paper describes the algorithm conceptually and mathematically, but does not include a clearly labeled pseudocode or algorithm block.
Open Source Code No The paper does not provide an explicit statement about open-source code release or a link to a code repository for the described methodology.
Open Datasets Yes Multi-view Car Detection: To train a multi-view car detector, we collected images of 128 Frontal, 100 Rear, 103 Left, and 103 Right car views. These were resized to 41 70 pixels. The multi-view car detector was evaluated on the USC car dataset [6], which consists of 197 color images of size 480 640, containing 410 instances of cars in different views. For the detection of traffic signs, we extracted 1, 159 training examples from the first set of the Summer traffic sign dataset [7].
Dataset Splits No The paper mentions training and testing data but does not explicitly specify a validation set split or detailed percentages/counts for the data partitioning for training, validation, and testing.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments (e.g., GPU/CPU models, memory).
Software Dependencies No The paper mentions algorithms and features used (e.g., ECBoost, cost-sensitive Boosting, integral channel features) but does not specify versions for any software dependencies or libraries.
Experiment Setup Yes All binary cascade detectors were learned with a combination of the ECBoost algorithm of [14] and the cost-sensitive Boosting method of [18]. Following [2], all cascaded detectors used integral channel features and trees of depth two as weak learners. The training parameters were set to η = 0.02, D = 0.95, FP = 10 6 and the training set was bootstrapped whenever the false positive rate dropped below 90%.