RAISE has a paper accepted at ACIVS 2026 (International Conference on Advanced Concepts for Intelligent Vision Systems), to be held in Okinawa, Japan, in December 2026: "Swin-AQI: Depth-Aware Vision Transformer with Physics-Informed Learning for Particulate Matter Estimation from Ground Imagery." Authors: Van-Truong-Thinh Nguyen, Thanh-Hai Tran and Dr. Xuan-Bach Le. Street-level imagery is a cheap and widely available complement to air-quality monitoring stations. Swin-AQI leans on the underlying physics: airborne particles scatter light along the line of sight, so distant buildings and the horizon lose contrast as particulate loading rises. The model steers post-backbone attention with a depth-coherence bias derived from MiDaS, encodes the hour of capture with learnable Fourier features, and constrains predictions with Beer–Lambert and Koschmieder terms. On 1,977 daylight images from Ho Chi Minh City, the model reaches R² = 0.80 for PM2.5 and R² = 0.76 for PM10, outperforming CNN and transformer baselines on the same images and labels. The authors are explicit about the limits: this is a within-distribution result, and the model is a single-city estimator for moderate pollution levels. Paper: https://lexuanbach.github.io/publication/ACIVS2026b.pdf Supplement: https://lexuanbach.github.io/publication/ACIVS2026b-supplement.pdf Code: https://github.com/ZleeMb0334/acivs2026_swin-aqi Dataset: https://kaggle.com/datasets/thnhnguynvntrng/swin-aqi-dataset