Yuchang Jiang

I'm a PhD student in the EcoVision Lab, Department of Mathematical Modeling and Machine Learning, University of Zurich, supervised by Jan Dirk Wegner and Konrad Schindler. My research focuses on computer vision and remote sensing, with an emphasis on vegetation parameter estimation, forest type mapping, and imbalanced data challenges. I am driven by a passion for applying deep learning techniques to tackle scientific challenges. I have interned in Google DeepMind, mainly working with Maxim Neumann and Dan Morris.

I earned my master's degree from ETH Zurich and my bachelor's degree from The Hong Kong Polytechnic University. During my master's studies, I specialized in computer vision and deep learning, focusing on multi-object tracking and remote sensing challenges.

Email  /  Github  /  Linkedin  /  Scholar

News

• March 2026: our paper "🍀 MIRANDA: MId-feature RANk-adversarial Domain Adaptation toward climate change-robust ecological forecasting with deep learning" is accepted at CVPR EarthVision 2026! check the project page
• Feb 2026: the preprint of our work "Tree crop mapping of South America reveals links to deforestation and conservation" is now available: paper link
• Nov 2025: our work "Natural Forests of the World" is online: Nature Scientific Data paper with a Google Rearch blog.
• Aug 2025: I'll give an oral presentation of our work, 🌳 Not every tree is a forest at IGARSS 2025 in Brisbane, Australia.
• June 2025: I'll present our tree crop work at Living Planet Symposium in Vienna, Austria.
• May 2025: Updates about JEO: MTSViT is open-sourced in it.
• 6 May 2025: I'll give a presentation about 🌳 FORTY benchmark at ESA-NASA International Workshop on AI Foundation Model for EO in Frascati, Italy.
• Mar 2025: one paper accepted at IEEE International Geoscience and Remote Sensing Symposium, 2025
• Feb 2025: 🌏JEO (a ML package for remote sensing and earth observation in JAX) is online! More codes will be open soon.

Data Download

Research

I'm interested in computer vision, deep learning, and their applications in remote sensing.

While phenology modelling has traditionally relied on mechanistic approaches, deep learning methods have recently been proposed as flexible, data-driven alternatives with often superior performance. However, mechanistic models tend to outperform deep networks when data distribution shifts are induced by climate change. Domain Adaptation (DA) techniques could help address this limitation. Yet, unlike standard DA settings, climate change induces a temporal continuum of domains and involves both a covariate and label shift, with warmer records and earlier start of spring. To tackle this challenge, we introduce Mid-feature Rank-adversarial Domain Adaptation (MIRANDA). We apply adversarial regularization to intermediate features. Moreover, instead of a binary domain-classification objective, we employ a rank-based objective that enforces year-invariance in the learned meteorological representations.

Monitoring tree crop expansion is vital for zero-deforestation policies like the European Union’s Regulation on Deforestation-free Products (EUDR). However, these efforts are hindered by a lack of highresolution data distinguishing diverse agricultural systems from forests. Here, we present the first 10m-resolution tree crop map for South America, generated using a multi-modal, spatio-temporal deep learning model trained on satellite imagery time series. While the map is experimental and carries known uncertainties, it offers a new perspective on the distribution of complex tree crop agricultural system

Identifying natural forests, which serve as critical biodiversity hotspots and major carbon sinks, is particularly valuable. We developed a novel global natural forest map for 2020 at 10 m resolution. This map can support initiatives like the European Union's Deforestation Regulation (EUDR) and other forest monitoring or conservation efforts that require a comprehensive baseline for monitoring deforestation and degradation.

ForTy is a new global-scale, multi-modal, and multi-temporal benchmark dataset designed for advancing global FORest TYpes mapping. It comprises 200,000 time series of image patches, each including Sentinel-2, Sentinel-1, climate, and elevation data. The dataset features per-pixel annotations that distinguish between three key forest types (natural forest, planted forest, tree crops).

We propose a new way to address high-resolution above-ground biomass estimation, by leveraging both high-resolution (HR) satellite observations and existing low-resolution (LR) biomass products. We cast this problem as Guided Super-Resolution, aiming at upsampling LR biomass maps (sources) from 100 to 10 m resolution, using auxiliary HR co-registered satellite images (guides).

UVOTE integrates recent advances in probabilistic deep learning with an ensemble approach for imbalanced regression. We replace traditional regression losses with negative log-likelihood, which also predicts sample-wise aleatoric uncertainty.

We generate annual, countrywide vegetation height maps at a 10-m ground sampling distance for the years 2017–2020 based on Sentinel-2 satellite imagery and deep learning.

We propose Semantic Spray (Semspray), a Virtual Reality (VR) application that provides users with intuitive and handy tools to produce semantic information on as-is 3D spatial data (mesh) of buildings.

We develop a deep learning framework that learns a joint satellite embedding to fuse MODIS, VIIRS and S1-SAR satellite data for lake icce monitoring.

We propose an approach to track multiple drones in a roughly synchronized static camera network with unknown camera poses.

We construct a working pipeline for 3d scooer player position tracking, including multi-object tracking in each camera view and multi-camera association.

Misc

I enjoy sports during feel time, mainly badmiton, squash, and swimming. My badminton team in BCZA is always looking for league palyers, especially female players for 2/3. Liga. Feel free to contact me if you're interested 🏸 🏸 🏸

Thank you for the template Jon Barron.