Image-Adaptive YOLO for Object Detection in Adverse Weather Conditions
Authors: Wenyu Liu, Gaofeng Ren, Runsheng Yu, Shi Guo, Jianke Zhu, Lei Zhang1792-1800
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The experimental results are very encouraging, demonstrating the effectiveness of our proposed IA-YOLO method in both foggy and low-light scenarios. |
| Researcher Affiliation | Collaboration | Wenyu Liu1,2*, Gaofeng Ren 3, Runsheng Yu 4, Shi Guo 5, Jianke Zhu 1,2 , Lei Zhang 3,5 1 Colleage of Computer Science and Technology, Zhejiang University 2 Alibaba-Zhejiang University Joint Institute of Frontier Technologies 3 DAMO Academy, Alibaba Group 4 The Hong Kong University of Science and Technology 5 The Hong Kong Polytechnic University |
| Pseudocode | Yes | Algorithm 1: Image-Adaptive YOLO training procedure |
| Open Source Code | Yes | The source code can be found at https://github.com/wenyyu/ImageAdaptive-YOLO. |
| Open Datasets | Yes | We build upon the classic VOC dataset (Everingham et al. 2010) a VOC_Foggy dataset according to the atmospheric scattering model (Narasimhan and Nayar 2002). Moreover, RTTS (Li et al. 2018) is a relatively comprehensive real-world dataset available in foggy conditions... PSCAL VOC (Everingham et al. 2010) and the relatively comprehensive low-light detection dataset Ex Dark (Loh and Chan 2019) both contain ten categories of objects... |
| Dataset Splits | No | The paper describes how training and test sets (e.g., VOC_norm_trainval, VOC_norm_test) are formed. However, it does not provide specific details on a separate validation set split (e.g., percentages or counts) needed to reproduce the data partitioning for validation. |
| Hardware Specification | Yes | We use Tensorflow for our experiments and run it on the Tesla V100 GPU. |
| Software Dependencies | No | The paper mentions 'Tensorflow' but does not specify a version number or any other software dependencies with version numbers. |
| Experiment Setup | Yes | The backbone network for all experiments is Darknet-53. During training, We randomly resize the image to (32N 32N), where N [9, 19]. Moreover, the data augmentation methods like image flipping, cropping and transformation are applied to expand the training dataset. Our IA-YOLO model is trained by the Adam optimizer (Kingma and Ba 2014) with 80 epochs. The starting learning rate is 10 4 and the batch size is 6. |