Upload folder using huggingface_hub

README.md

@@ -0,0 +1,21 @@
+
+---
+pipeline_tag: image-to-image
+tags:
+  - ESRGAN
+  - super-resolution
+  - enhancer
+---
+
+# Finegrain Image Enhancer (ESRGAN-based)
+
+This model enhances image quality using ESRGAN and custom ControlNet/LoRA techniques.
+
+## Usage
+
+```python
+from huggingface_hub import InferenceClient
+
+client = InferenceClient("your-username/finegrain-image-enhancer", token="hf_xxx")
+result = client.post(json={"inputs": {"image": "<base64 image>"}})
+```

checkpoints/4x-UltraSharp.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7bf14a0e758e545859068257e7138a0536f5c8ed6643a0fd66d9e72198a60b64
+size 35

checkpoints/JuggernautNegative-neg.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73989087e53cedffc98a052820411d256a1e140d8e3845ff736b57961f82130d
+size 28

checkpoints/SDXLrender_v2.0.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4522278873e3b6b0a8c064218bba4524be2e50911e66e187c9193019aeb46e27
+size 23

checkpoints/autoencoder_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:592ff7a3041445f7ff7e599894dcb53d1c23aa30e00bc1201f649e6bed7f1c00
+size 51

checkpoints/controlnet_tile_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b60f28e776a8d38e2eae2efe01b491f542eec0a0ce4fd3709a5668b4a46cd5db
+size 37

checkpoints/more_details.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4522278873e3b6b0a8c064218bba4524be2e50911e66e187c9193019aeb46e27
+size 23

checkpoints/text_encoder_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec6e6f9bc20a82042f59928440f5ff4ecd921122a59cf351f93bd991d3d1a36e
+size 52

checkpoints/unet_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fca60ff0bfa80e223bd08941f24117784ae87d2eaffebb55b54248fd32822a6
+size 44

config.json

@@ -0,0 +1,6 @@
+
+{
+  "pipeline_tag": "image-to-image",
+  "model_type": "esrgan",
+  "base_model": "philz1337x/upscaler"
+}

enhancer.py

@@ -0,0 +1 @@
+<copied from uploaded enhancer.py>

esrgan_model.py

@@ -0,0 +1 @@
+<copied from uploaded esrgan_model.py>

inference.py

@@ -0,0 +1,42 @@
+
+from pathlib import Path
+import torch
+from PIL import Image
+import base64
+import io
+from enhancer import ESRGANUpscaler, ESRGANUpscalerCheckpoints
+
+checkpoints = ESRGANUpscalerCheckpoints(
+    esrgan=Path("checkpoints/4x-UltraSharp.pth")
+)
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+dtype = torch.bfloat16 if torch.cuda.is_bf16_supported() else torch.float32
+
+enhancer = ESRGANUpscaler(
+    checkpoints=checkpoints,
+    device=device,
+    dtype=dtype
+)
+
+def inference(inputs: dict) -> dict:
+    if "image" not in inputs:
+        return {"error": "No image provided"}
+
+    image_data = inputs["image"]
+    if image_data.startswith("data:image"):
+        image_data = image_data.split(",")[1]
+    image_bytes = base64.b64decode(image_data)
+    input_image = Image.open(io.BytesIO(image_bytes)).convert("RGB")
+
+    enhanced_image = enhancer.upscale(input_image)
+
+    buf = io.BytesIO()
+    enhanced_image.save(buf, format="PNG")
+    b64 = base64.b64encode(buf.getvalue()).decode("utf-8")
+
+    return {
+        "enhanced_image": b64,
+        "original_size": input_image.size,
+        "enhanced_size": enhanced_image.size
+    }

requirements.txt

@@ -0,0 +1,8 @@
+
+git+https://github.com/finegrain-ai/refiners@cfe8b66ba4f8a906583850ac25e9e89cb83a44b9
+numpy<2.0.0
+pillow>=10.4.0
+pillow-heif>=0.18.0
+torch>=2.0.0
+transformers>=4.21.0
+huggingface-hub>=0.16.0