Transfer Learning from Synthetic Data in the Camera Pose Estimation Problem

Charco, Jorge L.; Sappa, Ángel D.; Vintimilla, Boris X.; Velesaca, Henry O.

doi:10.5220/0009167604980505

Cited by 5 publications

(1 citation statement)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to overcome the limitation of obtaining real-world images with haze, 3D models of different scenarios are required in order to simulate realistic haze image datasets. It should be mentioned that the usage of 3D virtual environments to generate a dataset of synthetic images has already been considered for tackling different computer vision problems for instance object recognition (e.g., pedestrians (Fabbri et al, 2021), vehicles (Tang et al, 2019)), camera calibration (Charco et al, 2018;Charco et al, 2020;Charco et al, 2021), just to mention a few. In the current work, a similar strategy is followed to address the problem of image haze removal.…”

Section: Paired Real Images (Clear/haze)mentioning

confidence: 99%

Domain Adaptation in Image Dehazing: Exploring the Usage of Images From Virtual Scenarios

2022

Proceedings of the 16th International Conference on Computer Graphics, Visualization, Computer Vision and Image Processing (CGV

View full text Add to dashboard Cite

This work presents a novel domain adaptation strategy for deep learning-based approaches to solve the image haze removal problem. Firstly, a large set of synthetic images is generated by using a realistic 3D graphic simulator; these synthetic images contain different densities of haze, which are used for training the model that is later adapted to any real scenario. The adaptation process requires just a few images to fine-tune the model parameters. The proposed strategy allows overcoming the limitation of training a given model with few images. In other words, the proposed strategy implements the adaptation of a haze removal model trained with synthetic images to real scenarios. It should be noticed that it is quite difficult, if not impossible, to have large sets of pairs of real-world images (with and without haze) to train in a supervised way haze removal algorithms. Experimental results are provided showing the validity of the proposed domain adaptation strategy.

show abstract