Muthukumaran commited on
Commit
24a7492
1 Parent(s): 453e361

impact-comms-update

Browse files

changes from the NASA-IMPACT O&C team

Files changed (1) hide show
  1. README.md +6 -6
README.md CHANGED
@@ -18,7 +18,7 @@ metrics:
18
  ---
19
 
20
  ### Model and Inputs
21
- The pretrained [Prithvi-100m](https://huggingface.co/ibm-nasa-geospatial/Prithvi-100M/blob/main/README.md) parameter model is finetuned to detect Burn Scars on HLS data from the [HLS Burn Scar Scenes dataset](https://huggingface.co/datasets/ibm-nasa-geospatial/hls_burn_scars). This dataset includes input tiles of 512x512x6, where 512 is the height and width and 6 is the number of bands. The bands are:
22
 
23
  1. Blue
24
  2. Green
@@ -27,19 +27,19 @@ The pretrained [Prithvi-100m](https://huggingface.co/ibm-nasa-geospatial/Prithvi
27
  5. SWIR 1
28
  6. SWIR 2
29
  ![](burn_scar.png)
30
- It is important to point out that the HLS Burn Scar Scenes dataset includes a single timestep, while the Prithvi-100m was pretrained with 3 timesteps. This highlights the flexibility of this model to adapt to different downstream tasks and requirements.
31
 
32
  ### Code
33
- Code for Finetuning is available through [github](https://github.com/NASA-IMPACT/hls-foundation-os/tree/main/fine-tuning-examples)
34
 
35
- Configuration used for finetuning is available through [config](https://github.com/NASA-IMPACT/hls-foundation-os/blob/main/fine-tuning-examples/configs/firescars_config.py
36
  ).
37
 
38
  ### Results
39
- The experiment by running the mmseg stack for 50 epochs using the above config led to an IoU of **0.72** on the burn scar class and **0.96** overall accuracy. It is noteworthy that this leads to a resonably good model, but further developement will most likely improve performance.
40
 
41
  ### Inference and demo
42
- The github repo includes an inference script that allows to run the burn scar model for inference on HLS images. These input have to be geotiff format, including the channels described above (Blue, Green, Red, Narrow NIR, SWIR, SWIR 2) in reflectance units [0-1]. There is also a **demo** that leverages the same code **[here](https://huggingface.co/spaces/ibm-nasa-geospatial/Prithvi-100M-Burn-scars-demo)**.
43
 
44
  ### Citation
45
 
 
18
  ---
19
 
20
  ### Model and Inputs
21
+ The pre-trained [Prithvi-100m](https://huggingface.co/ibm-nasa-geospatial/Prithvi-100M/blob/main/README.md) parameter model is fine-tuned to detect burn scars on HLS data from the [HLS burn scar scenes dataset](https://huggingface.co/datasets/ibm-nasa-geospatial/hls_burn_scars). This dataset includes input tiles of 512x512x6, where 512 is the height and width and 6 is the number of bands. The bands are:
22
 
23
  1. Blue
24
  2. Green
 
27
  5. SWIR 1
28
  6. SWIR 2
29
  ![](burn_scar.png)
30
+ It is important to point out that the HLS burn scar scenes dataset includes a single timestep, while the Prithvi-100m was pretrained with three timesteps. The difference highlights the flexibility of this model to adapt to different downstream tasks and requirements.
31
 
32
  ### Code
33
+ Code for fine-tuning is available through [Github](https://github.com/NASA-IMPACT/hls-foundation-os/tree/main/fine-tuning-examples)
34
 
35
+ Configuration used for fine-tuning is available through [config](https://github.com/NASA-IMPACT/hls-foundation-os/blob/main/fine-tuning-examples/configs/firescars_config.py
36
  ).
37
 
38
  ### Results
39
+ The experiment conducted by running the mmseg stack for 50 epochs using the above config led to an IoU of **0.72** on the burn scar class and **0.96** overall accuracy. It is noteworthy that this leads to a reasonably good model, but further developement will most likely improve performance.
40
 
41
  ### Inference and demo
42
+ The github repo includes an inference script that allows to run the burn scar model for inference on HLS images. These inputs have to be in geotiff format, including the channels described above (Blue, Green, Red, Narrow NIR, SWIR, SWIR 2) in reflectance units [0-1]. A **demo** that leverages the same code can be found **[here](https://huggingface.co/spaces/ibm-nasa-geospatial/Prithvi-100M-Burn-scars-demo)**.
43
 
44
  ### Citation
45