first commit

app.py

@@ -0,0 +1,52 @@
+# app.py
+import gradio as gr
+import torch
+from basicsr.archs.rrdbnet_arch import RRDBNet
+from realesrgan import RealESRGANer
+import cv2
+import os
+
+# Fungsi upscaling Anda yang sudah ada
+def upscale_image_gradio(input_image):
+    # Pastikan direktori models ada
+    model_path = 'models/RealESRGAN_x4plus.pth'
+    if not os.path.exists(model_path):
+        # Anda perlu memastikan model .pth diunduh ke folder 'models'
+        # atau Anda bisa mengunduhnya secara programatik jika tidak ada
+        raise FileNotFoundError(f"Model file not found at {model_path}. Please download it.")
+
+    model = RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64,
+                    num_block=23, num_grow_ch=32, scale=4)
+    upsampler = RealESRGANer(
+        scale=4,
+        model_path=model_path,
+        model=model,
+        tile=0, # Atur tile sesuai kebutuhan memori GPU/CPU Anda
+        tile_pad=10,
+        pre_pad=0,
+        half=False # Atur True jika Anda menggunakan GPU yang mendukung half-precision (FP16)
+    )
+
+    # input_image dari Gradio biasanya adalah numpy array (dari PIL Image yang di-load)
+    # Pastikan formatnya sesuai (BGR untuk OpenCV jika model Anda dilatih dengan itu)
+    # Gradio mengembalikan RGB, Real-ESRGAN/OpenCV biasanya butuh BGR
+    img_bgr = cv2.cvtColor(input_image, cv2.COLOR_RGB2BGR)
+
+    output_bgr, _ = upsampler.enhance(img_bgr, outscale=4)
+
+    # Konversi kembali ke RGB untuk ditampilkan oleh Gradio
+    output_rgb = cv2.cvtColor(output_bgr, cv2.COLOR_BGR2RGB)
+    
+    return output_rgb
+
+# Definisikan antarmuka Gradio
+iface = gr.Interface(
+    fn=upscale_image_gradio,
+    inputs=gr.Image(type="numpy", label="Upload Image"),
+    outputs=gr.Image(type="numpy", label="Upscaled Image"),
+    title="Real-ESRGAN Image Upscaler",
+    description="Upload an image to upscale it using Real-ESRGAN x4plus model."
+)
+
+if __name__ == "__main__":
+    iface.launch() # Untuk menjalankan lokal

models/RealESRGAN_x4plus.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fa0d38905f75ac06eb49a7951b426670021be3018265fd191d2125df9d682f1
+size 67040989

real_esrgan_utils.py

@@ -0,0 +1,22 @@
+import torch
+from basicsr.archs.rrdbnet_arch import RRDBNet
+from realesrgan import RealESRGANer
+import cv2
+
+def upscale_image(input_path, output_path):
+    model = RRDBNet(num_in_ch=3, num_out_ch=3, num_feat=64,
+                    num_block=23, num_grow_ch=32, scale=4)
+    upsampler = RealESRGANer(
+        scale=4,
+        model_path='models/RealESRGAN_x4plus.pth',
+        model=model,
+        tile=0,
+        tile_pad=10,
+        pre_pad=0,
+        half=False
+    )
+
+    img = cv2.imread(input_path, cv2.IMREAD_COLOR)
+    output, _ = upsampler.enhance(img, outscale=4)
+    cv2.imwrite(output_path, output)
+    return output_path

requirements.txt

@@ -0,0 +1,6 @@
+gradio
+realesrgan==0.3.0
+torch==2.1.0
+torchvision==0.16.0
+opencv-python
+numpy<2