Add application file

.gitignore

@@ -0,0 +1,7 @@
+.env
+.venv
+
+# ignore all __pycache__ file
+__pycache__/
+
+*.pyc

Dockerfile

@@ -0,0 +1,14 @@
+FROM python:3.9.4
+
+RUN useradd -m -u 1000 haydpw
+USER haydpw
+ENV PATH="/home/user/.local/bin:$PATH"
+
+COPY --chown=user . .
+
+WORKDIR /
+
+RUN pip install --no-chace-dir --upgrade -r requirements.txt
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]
+

README.md

@@ -1,11 +1,2 @@
----
-title: Skin Segmentation And Scoring
-emoji: 🐠
-colorFrom: purple
-colorTo: indigo
-sdk: docker
-pinned: false
-license: mit
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# skin-segmentation-api
+Semantic skin segmentation for Paragon beauty app

api/roboflow_inference_client.py

@@ -0,0 +1,13 @@
+from inference_sdk import InferenceHTTPClient
+import dotenv
+import os
+
+
+
+def create_client(api_key):
+    client = InferenceHTTPClient(
+        api_url="https://classify.roboflow.com",
+        api_key=api_key
+    )
+    return client
+    

exceptions/NotFaceError.py

@@ -0,0 +1,3 @@
+class NotFaceError(Exception):
+    def __init__(self, *args: object) -> None:
+        super().__init__(*args)

main.py

@@ -0,0 +1,84 @@
+from typing import Union
+import dotenv
+import traceback
+import json
+import io
+import os
+import base64
+from fastapi import FastAPI, File, HTTPException, UploadFile, Response
+import models.face_classifier as classifier
+from fastapi.middleware.cors import CORSMiddleware
+from PIL import Image
+from rembg import remove
+from utils.helpers import image_to_base64, calculate_mask_area
+
+
+dotenv.load_dotenv()
+
+app = FastAPI()
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+
+# CLIENT = create_client(os.getenv("ROBOFLOW_API_KEY"))
+# model = FaceClassifierModel(client=CLIENT)
+
+model = classifier.FaceSegmentationModel()
+
+
+
+@app.post("/segment/", summary="Classify skin type based on image given",tags=["Classify"])
+async def predict_image(file: UploadFile = File(...)):
+    try:
+        # Menangani file yang diunggah
+        image_file = await file.read()
+        pil_image= Image.open(io.BytesIO(image_file)).convert("RGB")
+        
+        # resize image to 512x512
+        pil_image = pil_image.resize((512, 512))
+        
+        image_bg_removed= remove(pil_image, bgcolor=(0,0,255,255))
+        
+        # Memanggil metode classify untuk melakukan klasifikasi
+        results = model.infer(image_bg_removed)
+        
+        print(len(results))
+        
+        background_element = next((element for element in results if element['label'] == 'background'), None)
+        
+        if background_element:
+            background_area = calculate_mask_area(background_element['mask'])
+        else:
+            background_area = 0
+        
+        # change the mask to base64 and calculate the score
+        for i in range(len(results)):
+            results[i]["mask"] = image_to_base64(results[i]["mask"])
+            if results[i]["label"] == "background":
+                continue
+            mask_area = calculate_mask_area(results[i]["mask"])
+            score = mask_area / (512 * 512 - background_area)
+            results[i]["score"] = score
+            
+            
+        # add original image base 64 as original image:
+        image_bg_removed = image_bg_removed.convert("RGB")
+        
+        response = {
+            "original_image": image_to_base64(image_bg_removed),
+            "segmentation_results":results
+        }
+        
+        # Kembalikan hasil klasifikasi
+        return Response(content=json.dumps(response), status_code=200)
+    
+    except Exception as e:
+        # Mendapatkan stack trace
+        error_traceback = traceback.format_exc()
+        raise HTTPException(status_code=500, detail=f"An error occurred: {str(e)}")

models/efficientnet_face_detection.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5317ef37d26a7eb4803afbbfd9c806973a208ecb34a13b3933d2395ac720d84f
+size 16678304

models/face_classifier.py

@@ -0,0 +1,98 @@
+from keras._tf_keras.keras.models import load_model
+import keras
+import warnings
+import traceback
+import cv2
+import sys
+import tensorflow as tf
+import numpy as np
+import exceptions
+import os
+from PIL import Image
+from exceptions.NotFaceError import NotFaceError
+from inference_sdk import InferenceHTTPClient
+from transformers import pipeline, SegformerForSemanticSegmentation, SegformerImageProcessor, SegformerFeatureExtractor
+
+def warning_with_traceback(message, category, filename, lineno, file=None, line=None):
+    log = file if hasattr(file,'write') else sys.stderr
+    traceback.print_stack(file=log)
+    log.write(warnings.formatwarning(message, category, filename, lineno, line))
+    
+warnings.showwarning = warning_with_traceback
+
+
+@keras.saving.register_keras_serializable()
+class CustomPreprocessingLayer(tf.keras.layers.Layer):
+    def __init__(self, input_shape, **kwargs):
+        self.input_shape = input_shape
+        super(CustomPreprocessingLayer, self).__init__(**kwargs)
+        
+    def build(self, input_shape):
+        pass  # No trainable weights to build
+
+    def call(self, image_matrix):
+        image = tf.convert_to_tensor(image_matrix, dtype=tf.int32)
+        image = tf.image.resize(image, [self.input_shape[0], self.input_shape[1]])
+        return image
+    def get_config(self):
+        config = super(CustomPreprocessingLayer, self).get_config()
+        config.update({'input_shape': self.input_shape})
+        return config
+    
+    @classmethod
+    def from_config(cls, config):
+        return cls(**config)
+
+class FaceClassifierModel:
+    def __init__(self,  client:InferenceHTTPClient, image_size=224, batcb_size=16):
+        self.model = load_model("./models/efficientnet_face_detection.h5")
+        self.image_size = image_size
+        self.batch_size = batcb_size
+        self.seed = 42
+        self.client = client
+    
+    async def classify(self, image_bytes: str, confidence_threshold=0.5):
+        tf.random.set_seed(self.seed)
+        nparr = np.frombuffer(image_bytes, np.uint8)
+        
+        # Dekode array NumPy menjadi citra OpenCV
+        image = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
+        image = cv2.resize(image, [self.image_size, self.image_size])
+        image_expanded = tf.expand_dims(image,axis=0)
+        image_batch = tf.data.Dataset.from_tensor_slices(image_expanded).batch(self.batch_size)
+        pred = self.model.predict(image_batch)
+        if pred[0][0] <= confidence_threshold:
+            raise NotFaceError("Ini bukan wajah")
+        
+        # lanjut klasifikasi muka
+        result = await self.client.infer_async(image, model_id="skinclassification-kyxvj/1")
+        result["face_confidence"] = float(pred[0][0])
+        return result
+    
+    
+class FaceSegmentationModel:
+    def __init__(self):
+        model_checkpoint = os.path.join("models","segformer-b0-finetuned-segments-skin-outputs", "checkpoint-1640")
+        self.model = SegformerForSemanticSegmentation.from_pretrained(model_checkpoint, local_files_only=True)
+        self.image_processor = SegformerImageProcessor.from_pretrained(model_checkpoint, local_files_only=True)
+        self.pipeline = pipeline("image-segmentation", model=self.model, image_processor=self.image_processor)
+    
+    def infer(self, image:Image.Image):
+        '''
+        Infer the input image. it will return list of {'score', 'label', and 'mask'}
+        
+        Example:
+            [{'score': None,
+            'label': 'background',
+            'mask': <PIL.Image.Image image mode=L size=500x500>},
+            {'score': None,
+            'label': 'acne',
+            'mask': <PIL.Image.Image image mode=L size=500x500>},
+            {'score': None,
+            'label': 'dry',
+            'mask': <PIL.Image.Image image mode=L size=500x500>}]
+        '''
+        results = self.pipeline(image)
+        return results
+        
+        

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1620/config.json

@@ -0,0 +1,82 @@
+{
+  "_name_or_path": "nvidia/mit-b0",
+  "architectures": [
+    "SegformerForSemanticSegmentation"
+  ],
+  "attention_probs_dropout_prob": 0.0,
+  "classifier_dropout_prob": 0.1,
+  "decoder_hidden_size": 256,
+  "depths": [
+    2,
+    2,
+    2,
+    2
+  ],
+  "downsampling_rates": [
+    1,
+    4,
+    8,
+    16
+  ],
+  "drop_path_rate": 0.1,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.0,
+  "hidden_sizes": [
+    32,
+    64,
+    160,
+    256
+  ],
+  "id2label": {
+    "0": "background",
+    "1": "acne",
+    "2": "dry",
+    "3": "oily"
+  },
+  "image_size": 224,
+  "initializer_range": 0.02,
+  "label2id": {
+    "acne": 1,
+    "background": 0,
+    "dry": 2,
+    "oily": 3
+  },
+  "layer_norm_eps": 1e-06,
+  "mlp_ratios": [
+    4,
+    4,
+    4,
+    4
+  ],
+  "model_type": "segformer",
+  "num_attention_heads": [
+    1,
+    2,
+    5,
+    8
+  ],
+  "num_channels": 3,
+  "num_encoder_blocks": 4,
+  "patch_sizes": [
+    7,
+    3,
+    3,
+    3
+  ],
+  "reshape_last_stage": true,
+  "semantic_loss_ignore_index": 255,
+  "sr_ratios": [
+    8,
+    4,
+    2,
+    1
+  ],
+  "strides": [
+    4,
+    2,
+    2,
+    2
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.41.2"
+}

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1620/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b113dd70ae08759ae88e63e3ba8c36c8d7ae14f0f73d62b37c3c62818427245a
+size 14886832

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1620/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cb5172747645a4de909559836444a27932598a7b60531576ea49e86a48a1289
+size 29886586

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1620/preprocessor_config.json

@@ -0,0 +1,39 @@
+{
+  "_valid_processor_keys": [
+    "images",
+    "segmentation_maps",
+    "do_resize",
+    "size",
+    "resample",
+    "do_rescale",
+    "rescale_factor",
+    "do_normalize",
+    "image_mean",
+    "image_std",
+    "do_reduce_labels",
+    "return_tensors",
+    "data_format",
+    "input_data_format"
+  ],
+  "do_normalize": true,
+  "do_reduce_labels": false,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.485,
+    0.456,
+    0.406
+  ],
+  "image_processor_type": "SegformerImageProcessor",
+  "image_std": [
+    0.229,
+    0.224,
+    0.225
+  ],
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 512,
+    "width": 512
+  }
+}

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1620/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:706d5e18e0bcc00af9508719f6801a358319a9b2bea62936270cc31440a2c5ee
+size 13990

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1620/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2813ff910e7d207a59baff21caaa804f1364972acd9b94863d3ee624dc0d1e3
+size 1064

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1620/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a40ebbe9f4a4f2e901367dcd786dad8a884dfe4f6555f2a55fbebf29372529d
+size 5112

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1640/config.json

@@ -0,0 +1,82 @@
+{
+  "_name_or_path": "nvidia/mit-b0",
+  "architectures": [
+    "SegformerForSemanticSegmentation"
+  ],
+  "attention_probs_dropout_prob": 0.0,
+  "classifier_dropout_prob": 0.1,
+  "decoder_hidden_size": 256,
+  "depths": [
+    2,
+    2,
+    2,
+    2
+  ],
+  "downsampling_rates": [
+    1,
+    4,
+    8,
+    16
+  ],
+  "drop_path_rate": 0.1,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.0,
+  "hidden_sizes": [
+    32,
+    64,
+    160,
+    256
+  ],
+  "id2label": {
+    "0": "background",
+    "1": "acne",
+    "2": "dry",
+    "3": "oily"
+  },
+  "image_size": 224,
+  "initializer_range": 0.02,
+  "label2id": {
+    "acne": 1,
+    "background": 0,
+    "dry": 2,
+    "oily": 3
+  },
+  "layer_norm_eps": 1e-06,
+  "mlp_ratios": [
+    4,
+    4,
+    4,
+    4
+  ],
+  "model_type": "segformer",
+  "num_attention_heads": [
+    1,
+    2,
+    5,
+    8
+  ],
+  "num_channels": 3,
+  "num_encoder_blocks": 4,
+  "patch_sizes": [
+    7,
+    3,
+    3,
+    3
+  ],
+  "reshape_last_stage": true,
+  "semantic_loss_ignore_index": 255,
+  "sr_ratios": [
+    8,
+    4,
+    2,
+    1
+  ],
+  "strides": [
+    4,
+    2,
+    2,
+    2
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.41.2"
+}

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1640/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7663fa19c0154a4aef1ba06568f117f16864bf54c25993d2aff19a245db8fae2
+size 14886832

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1640/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2490d4a61e40a1214274e1914d7f8076ec2242d14b456515d9591f184eac7bfb
+size 29886586

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1640/preprocessor_config.json

@@ -0,0 +1,39 @@
+{
+  "_valid_processor_keys": [
+    "images",
+    "segmentation_maps",
+    "do_resize",
+    "size",
+    "resample",
+    "do_rescale",
+    "rescale_factor",
+    "do_normalize",
+    "image_mean",
+    "image_std",
+    "do_reduce_labels",
+    "return_tensors",
+    "data_format",
+    "input_data_format"
+  ],
+  "do_normalize": true,
+  "do_reduce_labels": false,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.485,
+    0.456,
+    0.406
+  ],
+  "image_processor_type": "SegformerImageProcessor",
+  "image_std": [
+    0.229,
+    0.224,
+    0.225
+  ],
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 512,
+    "width": 512
+  }
+}

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1640/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9267e5c2da8f7e26291010326e3fae282f67edb5974de0985b158b51f4a83ef1
+size 13990

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1640/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83effdf75fb04c222304704de02ea25ba33e71443d005c74d835b27e9cb5e7d2
+size 1064

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-1640/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a40ebbe9f4a4f2e901367dcd786dad8a884dfe4f6555f2a55fbebf29372529d
+size 5112

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-800/config.json

@@ -0,0 +1,82 @@
+{
+  "_name_or_path": "nvidia/mit-b0",
+  "architectures": [
+    "SegformerForSemanticSegmentation"
+  ],
+  "attention_probs_dropout_prob": 0.0,
+  "classifier_dropout_prob": 0.1,
+  "decoder_hidden_size": 256,
+  "depths": [
+    2,
+    2,
+    2,
+    2
+  ],
+  "downsampling_rates": [
+    1,
+    4,
+    8,
+    16
+  ],
+  "drop_path_rate": 0.1,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.0,
+  "hidden_sizes": [
+    32,
+    64,
+    160,
+    256
+  ],
+  "id2label": {
+    "0": "background",
+    "1": "acne",
+    "2": "dry",
+    "3": "oily"
+  },
+  "image_size": 224,
+  "initializer_range": 0.02,
+  "label2id": {
+    "acne": 1,
+    "background": 0,
+    "dry": 2,
+    "oily": 3
+  },
+  "layer_norm_eps": 1e-06,
+  "mlp_ratios": [
+    4,
+    4,
+    4,
+    4
+  ],
+  "model_type": "segformer",
+  "num_attention_heads": [
+    1,
+    2,
+    5,
+    8
+  ],
+  "num_channels": 3,
+  "num_encoder_blocks": 4,
+  "patch_sizes": [
+    7,
+    3,
+    3,
+    3
+  ],
+  "reshape_last_stage": true,
+  "semantic_loss_ignore_index": 255,
+  "sr_ratios": [
+    8,
+    4,
+    2,
+    1
+  ],
+  "strides": [
+    4,
+    2,
+    2,
+    2
+  ],
+  "torch_dtype": "float32",
+  "transformers_version": "4.41.2"
+}

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-800/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d7053510fb0a0c8e6d2df90e8b01106c91745b9264c3949430b1e9089a9a478
+size 14886832

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-800/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa1f746da01695702a4bec54dfe2abe27a39edb5579004a1b970deb2fd27fcab
+size 29886586

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-800/preprocessor_config.json

@@ -0,0 +1,39 @@
+{
+  "_valid_processor_keys": [
+    "images",
+    "segmentation_maps",
+    "do_resize",
+    "size",
+    "resample",
+    "do_rescale",
+    "rescale_factor",
+    "do_normalize",
+    "image_mean",
+    "image_std",
+    "do_reduce_labels",
+    "return_tensors",
+    "data_format",
+    "input_data_format"
+  ],
+  "do_normalize": true,
+  "do_reduce_labels": false,
+  "do_rescale": true,
+  "do_resize": true,
+  "image_mean": [
+    0.485,
+    0.456,
+    0.406
+  ],
+  "image_processor_type": "SegformerImageProcessor",
+  "image_std": [
+    0.229,
+    0.224,
+    0.225
+  ],
+  "resample": 2,
+  "rescale_factor": 0.00392156862745098,
+  "size": {
+    "height": 512,
+    "width": 512
+  }
+}

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-800/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1b7af371e5ab3ec21ee88d4107bdd1fe2827f1d3f8e95f47292bafbf8b28622
+size 13990

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-800/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba02d6444725590e035a377e4bb8bb416bb1c21fb5d03ae7faa1cc51769398f5
+size 1064

models/segformer-b0-finetuned-segments-skin-outputs/checkpoint-800/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a40ebbe9f4a4f2e901367dcd786dad8a884dfe4f6555f2a55fbebf29372529d
+size 5112

models/segformer-b0-finetuned-segments-skin-outputs/runs/Jul10_13-17-20_44c7630ab4c4/events.out.tfevents.1720617441.44c7630ab4c4.317.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da95df5064431d7aa08d6994bea475a274e5008df1c46b239ad39aedf7d64de3
+size 397851

update_requirements.bat

@@ -0,0 +1,42 @@
+@echo off
+
+:: Check if --skip argument is present
+set skip_installation=false
+for %%i in (%*) do (
+    if "%%i"=="--skip" (
+        set skip_installation=true
+    )
+)
+
+:: Install packages from requirements.txt if not skipped and file exists
+if "%skip_installation%"=="false" (
+    if exist requirements.txt (
+        echo Installing packages from requirements.txt...
+        pip install -r requirements.txt
+    ) else (
+        echo requirements.txt not found. Skipping installation from file.
+    )
+) else (
+    echo Skipping installation from requirements.txt...
+)
+
+:: Remove --skip from the arguments
+set "args="
+for %%i in (%*) do (
+    if not "%%i"=="--skip" (
+        set "args=!args! %%i"
+    )
+)
+set args=%args:~1%
+
+:: Install additional packages passed as arguments
+if "%*" neq "" (
+    echo Installing additional packages: %*
+    pip install %*
+)
+
+:: Perbarui requirements.txt
+echo Updating requirements.txt...
+pip freeze > requirements.txt
+
+echo Done.

update_requirements.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+# Check if --skip argument is present
+skip_installation=false
+args=()
+
+for arg in "$@"; do
+    if [ "$arg" == "--skip" ]; then
+        skip_installation=true
+    else
+        args+=("$arg")
+    fi
+done
+
+# Install packages from requirements.txt if not skipped and file exists
+if [ "$skip_installation" == false ]; then
+    if [ -f requirements.txt ]; then
+        echo "Installing packages from requirements.txt..."
+        pip install -r requirements.txt
+    else
+        echo "requirements.txt not found. Skipping installation from file."
+    fi
+else
+    echo "Skipping installation from requirements.txt..."
+fi
+
+# Install additional packages passed as arguments
+if [ ${#args[@]} -ne 0 ]; then
+    echo "Installing additional packages: ${args[*]}"
+    pip install "${args[@]}"
+fi
+
+# Update requirements.txt
+echo "Updating requirements.txt..."
+pip freeze > requirements.txt
+
+echo "Done."

utils/helpers.py

@@ -0,0 +1,15 @@
+from PIL import Image
+import io
+import base64
+import numpy as np
+
+def image_to_base64(image: Image.Image) -> str:
+    buffered = io.BytesIO()
+
+    image.save(buffered, format="JPEG")
+    return base64.b64encode(buffered.getvalue()).decode("utf-8")
+
+def calculate_mask_area(mask: Image.Image) -> int:
+    mask_array = np.array(mask)
+    non_zero_pixels = np.count_nonzero(mask_array)
+    return non_zero_pixels