Announcement: this year, THREE challenges are being organized in the framework of PBVS'2023 workshop.
The fourth thermal image super-resolution (TISR) challenge consists of two tracks.
Track 1 follows the same setup as the third edition (two kinds of evaluations). Evaluation1, consists in generating a x4 super-resolution thermal image from the given down-sampled by 4 and noised set of images. Evaluation2, consist in generating a x2 SR image from a given mid-resolution set of images acquired with an Axis Q2901-E camera, and as an output, the corresponding high-resolution images, of the same scene, compared with the image acquired with a FLIR FC-632O camera.
Track 2 uses a new generated dataset, acquired with cross-spectral sensors (Balser and TAU2 camera). It consists of registered pairs of visible and thermal images of the same scene in 640x480 resolution. The goal of this track is to generate the x8 SR thermal image by using the HR visible image as a guidance for the LR thermal image.
More details and dataset in CodaLab page:
Electro-optical (EO) sensors that capture images in the visible spectrum such as RGB and grayscale images, have been most prevalent in the computer vision research area. However, other sensors such as synthetic aperture radar (SAR) can reproduce images from radar signals that in some cases could complement EO sensors when such sensors fail to capture significant information (i.e. weather condition, no visible light, etc)
An ideal automated target recognition system would be based on multi-sensor information to compensate for the shortcomings of either of the sensor-based platforms individually. However, it is currently unclear if/how using EO and SAR data together can improve the performance of automatic target recognition (ATR) systems. Thus, the motivation for this challenge is to understand if and how data from one modality can improve the learning process for the other modality and vice versa. Ideas from domain adaptation, transfer learning or fusion are welcomed to solve this problem.
In addition to target recognition, this challenge introduces out-of-distribution detection. A robust target recognition system would not only provide a labeled target but also a confidence score for the target. A low score would correspond to an out-of-distribution sample.
More details and dataset in CodaLab page:
We introduce a new sensor translation competition for 2023. Sensor translation algorithms allow for dataset augmentation and allows for the fusion of information from multiple sensors. Electro-optical (EO) and Synthetic Aperture Radar (SAR) sensors provide a unique environment for translation. The motivation for this challenge is to understand how if and how data from one modality can be translated to another modality. This competition challenges participants to design methods to translate aligned images from the SAR modality to the EO modality.
More details and dataset in CodaLab page: