NemesisAlm
commited on
Commit
•
98c74a0
1
Parent(s):
3c790ed
Update README.md
Browse files
README.md
CHANGED
@@ -15,24 +15,24 @@ metrics:
|
|
15 |
|
16 |
# clip-fine-tuned-satellite
|
17 |
|
18 |
-
This model is a fine-tuned version of [openai/clip-vit-base-patch32](https://huggingface.co/openai/clip-vit-base-patch32) on the blanchon/UC_Merced dataset
|
19 |
-
It achieves the following results on the test set
|
20 |
-
-Accuracy: 96.9%
|
21 |
The original CLIP model achieves 58.8% of accuracy.
|
22 |
|
23 |
## Model description
|
24 |
|
25 |
-
The model is a fine-tuned version of CLIP
|
26 |
30% of the parameters were retrained to achieve a significant increase in accuracy after only 2 epochs.
|
27 |
|
28 |
## Intended uses & limitations
|
29 |
|
30 |
-
The model is to be used to classify satellite images
|
31 |
It was trained on the UC_Merced dataset that comprises 21 classes: agricultural, airplane, baseballdiamond, beach, buildings, chaparral, denseresidential, forest, freeway, golfcourse, harbor, intersection, mediumresidential, mobilehomepark, overpass, parkinglot, river, runway, sparseresidential, storagetanks, tenniscourt
|
32 |
|
33 |
## Training and evaluation data
|
34 |
|
35 |
-
30% of the parameters trained
|
36 |
Evaluated against a test set of 420 images.
|
37 |
|
38 |
## Training procedure
|
|
|
15 |
|
16 |
# clip-fine-tuned-satellite
|
17 |
|
18 |
+
This model is a fine-tuned version of [openai/clip-vit-base-patch32](https://huggingface.co/openai/clip-vit-base-patch32) on the blanchon/UC_Merced dataset.\
|
19 |
+
It achieves the following results on the test set:\
|
20 |
+
-Accuracy: 96.9% \
|
21 |
The original CLIP model achieves 58.8% of accuracy.
|
22 |
|
23 |
## Model description
|
24 |
|
25 |
+
The model is a fine-tuned version of CLIP.\
|
26 |
30% of the parameters were retrained to achieve a significant increase in accuracy after only 2 epochs.
|
27 |
|
28 |
## Intended uses & limitations
|
29 |
|
30 |
+
The model is to be used to classify satellite images.\
|
31 |
It was trained on the UC_Merced dataset that comprises 21 classes: agricultural, airplane, baseballdiamond, beach, buildings, chaparral, denseresidential, forest, freeway, golfcourse, harbor, intersection, mediumresidential, mobilehomepark, overpass, parkinglot, river, runway, sparseresidential, storagetanks, tenniscourt
|
32 |
|
33 |
## Training and evaluation data
|
34 |
|
35 |
+
30% of the parameters trained.\
|
36 |
Evaluated against a test set of 420 images.
|
37 |
|
38 |
## Training procedure
|