Meta's deep learning model to estimate tree canopy height in high-resolution satellite imagery. A brief summary of the item is not available. Add a brief summary about the item.
Deep learning package
by
Item created: Dec 10, 2024 Item updated: Jan 9, 2025 Number of downloads: 5,065
Description
Monitoring tree canopy height is crucial for assessing forest health, biodiversity, and carbon sequestration potential, as it provides insights into forest structure and ecosystem dynamics. Lidar data, which is preferable for this use, isn't always available and other measurement methods can be labor-intensive and time-consuming, often limited to small areas. This model can be used to estimate tree canopy height given high-resolution satellite imagery where Lidar data isn't available.
This Deep Learning Package (DLPK) contains Meta's High-Resolution Canopy Height model. The model employs a vision transformer backbone pretrained using self-supervised learning on millions of high-resolution satellite images from around the globe. It then uses a convolutional decoder trained on a LiDAR-derived canopy height dataset to generate canopy height estimates, expressed in meters above ground. Use this model to automate the workflow for estimating tree canopy height from high-resolution satellite imagery over large areas.
Using the model
Follow the guide to use the model. Before using this model, ensure that the supported deep learning libraries are installed. For more details, check Deep Learning Libraries Installer for ArcGIS.
Fine-tuning the model
This model cannot be fine-tuned using ArcGIS tools.
Input
8-bit, 3-band high resolution (0.6 - 1 meter) satellite imagery.
Output
Classified raster with each pixel value representing the height of tree canopy in meters.
Applicable geographies
This model is expected to work well globally.
Model architecture
This model packages Meta's High Resolution Canopy Height model (Tolan et al., 2023).
Accuracy metrics
The model produces an average Mean Absolute Error (MAE) of 2.8 m and Mean Error (ME) of 0.6 m on NEON dataset.
Limitations
Lidar data can provide more accurate measurements where it is available and this model should only be used where such data isn't available. Prediction on regions with tree shadows, terrains with slope might have inconsistent results.
Sample results
Here are a few results from the model.
See this web scene for examples of 3D Trees derived from this model.
An in-depth description of the item is not available.
Layers
Tools
Tables
Basemap
Project Contents:
Solution Contents
Contents
Layers
Screenshots
Terms of Use
No special restrictions or limitations on using the item's content have been provided.
Details
Dashboard views: Desktop
Source:
Creating data in:
Published as:
Other Views:
Dependent items in the recycle bin
Applicable: 2d
Size: 2,724.803 MB
ID: dac319ec70f844dab2d152f32dfff399
Image Count: 0
Image Properties
Layer Drawing
Using tiles from a cache
Dynamically from data
Share
Owner
Folder
Categories
This item has not been categorized.
Credits (Attribution)
No acknowledgements.Meta
Comments (11)
Thank you for your response and sharing your email. Please check your email as soon as possible.
Dear vraj_deldev Please could you send me your email address so I can ask my question more precisely there.
Hi @Mohsen1373, you can send your queries on spathak@esri.com.
Unfortunately, no matter what I do, it doesn't work!!! I don't know where the problem is!
It also gives this warning: Warning 003387 Input raster cell size 1.000000 is not within the cell size range [11131.949079, 111319.490793] the Deep Learning model has been trained for. Results may not be optimal. Consider setting another cell size in the environment, or using a different input raster, or using another Deep Learning model. Please provide an explanation of the input type.
Hi, please double check if the cell size being considered by the tool is not in degree, and if it is in degree then the value should be within the guide's recommended range (in meters). (for example: 0.6 meters, when around 38 N latitude on earth, is approx. 0.000006852150189302777 degrees, which changes arcoss different latitude.). Hence for a 'input raster', only in geographic coordinate system, that should be mentioned in the cell size, instead 0.60. as the tool will consider 0.6 as 'degree' and not as 'meters'.
Hello, I select the input image as Ultra Small (with pixel size 0.25 meters). It gives an error!!! Even if I change the pixel size to 0.7 meters, it still gives an error. Please help me to fix this error.: A raster error has occurred. The messages that follow will provide more detail.
Hi, It is hard to tell with this info. But this might be due to the coordinate system of 'input raster'. For ease, reproject the raster into a projected coordinate system. Feel free to reach out if you face any other issue.