Learning Efficient Object Detection Models with Knowledge Distillation

Authors: Guobin Chen, Wongun Choi, Xiang Yu, Tony Han, Manmohan Chandraker

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct comprehensive empirical evaluation with different distillation configurations over multiple datasets including PASCAL, KITTI, ILSVRC and MS-COCO.
Researcher Affiliation Collaboration 1NEC Labs America 2University of Missouri 3University of California, San Diego
Pseudocode No The paper does not include any pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not provide an explicit statement about releasing source code or a link to a code repository.
Open Datasets Yes Datasets We evaluate our method on several commonly used public detection datasets, namely, KITTI [12], PASCAL VOC 2007 [11], MS COCO [6] and Image Net DET benchmark (ILSVRC 2014) [35].
Dataset Splits Yes Since KITTI and ILSVRC 2014 do not provide ground-truth annotation for test sets, we use the training/validation split introduced by [39] and [24] for analysis.
Hardware Specification No The paper mentions running experiments "on GPU" but does not specify any particular GPU models, CPU models, memory details, or other specific hardware specifications.
Software Dependencies No The paper does not explicitly list software dependencies with their specific version numbers.
Experiment Setup Yes We fix them [λ and γ] to be 1 and 0.5, respectively, throughout the experiments. For example, we use w0 = 1.5 for the background class and wi = 1 for all the others in experiments on the PASCAL dataset. ... ν is a weight parameter (set as 0.5 in our experiments).