Update README.md
Browse files
README.md
CHANGED
|
@@ -24,4 +24,14 @@ We will not share the dataset with any third party libraries. However since AIGC
|
|
| 24 |
|
| 25 |
Okay now let's get to the topic:
|
| 26 |
|
| 27 |
-
How can I download the model?
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
Okay now let's get to the topic:
|
| 26 |
|
| 27 |
+
Q: How can I download the model?
|
| 28 |
+
|
| 29 |
+
A: You should be able to directly download it from the 'Files' List, aka the "NSFW_Classifier.h5". You can then place it inside the directory of the Estallie-Inference.py file and start the code.
|
| 30 |
+
|
| 31 |
+
Q: How can I train my own model?
|
| 32 |
+
|
| 33 |
+
A: Go to Estallie-trainer.py, edit the Train and Validation folders and model name, then run the program. Remeber the default batch size is 4 so you might want to have more than 4 images per class. Also there is only two classes avaliable for this basic model.
|
| 34 |
+
|
| 35 |
+
Q: How are you supposed to interpret if an image is not NSFW or SFW such as Unethical...etc?
|
| 36 |
+
|
| 37 |
+
A: Good point. At this point we used seperate models to interpret. We also have yolov5 and Deepdanbooru but it's not fun anymore if they're trained for you. We have a backup program for automatically classifying subclasses, the "1c3a.oy" file, and we are trying to complete it.
|