The western Himalayas region is a key biodiversity hotspot and a habitat for several species. This region needs an accurate spatial assessment of tree canopy height for improved estimates of aboveground biomass, carbon sequestration, and related forest ecosystem services. In this work, we used a deep convolution neural networks model, U-Net, for estimating canopy heights using Global Ecosystem Dynamics Investigation Mission (GEDI) with multi-band Sentinel-2 datasets. We produced a 10 m canopy height map of the study region for 2020. Overall, we achieved a Root Mean Square Error (RMSE) of 7.52 m and a Mean Absolute Error (MAE) of 5.71 m on test set, outperforming existing global canopy height models in this topographically challenging region. Citation: Use of these data require citation of this dataset: Ahmad, A., Sastry, S., Dhakal, A., Khanal, S., Levering, A., Gilani, H., & Jacobs, N. (2025). WHiCH (Western Himalaya Canopy Height) Map 2020, Pakistan [Data set]. In International Journal of Applied Earth Observation and Geoinformation. Zenodo. https://doi.org/10.5281/zenodo.18056032 Original research article: Ahmad, A., Sastry, S., Dhakal, A., Khanal, S., Levering, A., Gilani, H. & Jacobs, N. (2025). Canopy Height Mapping in the Western Himalayas, Pakistan: A Deep Learning Approach using GEDI and Sentinel-2 Fusion. International Journal of Applied Earth Observation and Geoinformation. doi.org/10.1016/j.jag.2025.105030