Hugging Face's logo

MITLL
/
LADI-v2-classifier-small-reference

Image Classification
Transformers
Safetensors
bit
LADI
Aerial Imagery
Disaster Response
Emergency Management
Model card Files
xet
 Community
jeffliu-LL commited on
Commit
c6178de
·
verified ·
1 Parent(s): 91d17db

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +2 -2
README.md CHANGED
@@ -14,7 +14,7 @@ tags:
14
  # Model Card for MITLL/LADI-v2-classifier-small-reference
15
  LADI-v2-classifier-small-reference is based on [google/bit-50](https://huggingface.co/google/bit-50) and fine-tuned on the LADI v2_resized dataset. LADI-v2-classifier is trained to identify labels of interest to disaster response managers from aerial images.
16
 
17
- 🔴 __IMPORTANT__ ❗🔴 This model is the 'reference' version of the model, which is trained on 80% of the 10,000 available images. It is provided to facilitate reproduction of our paper and is not intended to be used in deployment.
18
 
19
  ## Model Details
20
 
@@ -34,7 +34,7 @@ The model architecture is based on Google's bit-50 model and fine-tuned on the L
34
  - trees_damage
35
  - water_any
36
 
37
- This 'reference' model is trained on the training split, which contains 8,000 images from 2015-2022. It is provided for the purpose of reproducing the results from the paper. The 'deploy' model is trained on the training, validation, and test sets, and contains 10,000 images from 2015-2023. We recommend that anyone who wishes to use this model in production use the deploy model.
38
 
39
  - **Developed by:** Jeff Liu, Sam Scheele
40
  - **Funded by:** Department of the Air Force under Air Force Contract No. FA8702-15-D-0001
 
14
  # Model Card for MITLL/LADI-v2-classifier-small-reference
15
  LADI-v2-classifier-small-reference is based on [google/bit-50](https://huggingface.co/google/bit-50) and fine-tuned on the LADI v2_resized dataset. LADI-v2-classifier is trained to identify labels of interest to disaster response managers from aerial images.
16
 
17
+ 🔴 __IMPORTANT__ ❗🔴 This model is the 'reference' version of the model, which is trained on 80% of the 10,000 available images. It is provided to facilitate reproduction of our paper and is not intended to be used in deployment. For deployment, see the [MITLL/LADI-v2-classifier-small](https://huggingface.co/MITLL/LADI-v2-classifier-small) and [MITLL/LADI-v2-classifier-large](https://huggingface.co/MITLL/LADI-v2-classifier-large) models, which are trained on the full LADI v2 dataset (all splits).
18
 
19
  ## Model Details
20
 
 
34
  - trees_damage
35
  - water_any
36
 
37
+ This 'reference' model is trained on the training split, which contains 8,000 images from 2015-2022. It is provided for the purpose of reproducing the results from the paper. The 'deploy' model is trained on the training, validation, and test sets, and contains 10,000 images from 2015-2023. We recommend that anyone who wishes to use this model in production use the main versions of the models [MITLL/LADI-v2-classifier-small](https://huggingface.co/MITLL/LADI-v2-classifier-small) and [MITLL/LADI-v2-classifier-large](https://huggingface.co/MITLL/LADI-v2-classifier-large).
38
 
39
  - **Developed by:** Jeff Liu, Sam Scheele
40
  - **Funded by:** Department of the Air Force under Air Force Contract No. FA8702-15-D-0001