Upload 12 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+V4EauuhVEw4.jpg filter=lfs diff=lfs merge=lfs -text

app.py

@@ -12,26 +12,42 @@ from transformers import BlipProcessor, BlipForConditionalGeneration
 title = "# 🗜️ CLaMP 3 - Multimodal & Multilingual Semantic Music Search"
 
 badges = """
-    <div style="text-align: center;">
-        <a href="#"><img src="https://img.shields.io/badge/CLaMP%203%20Homepage-Coming%20Soon-lightgrey?style=for-the-badge&logo=home-assistant" alt="Homepage"></a>
-        <a href="#"><img src="https://img.shields.io/badge/CLaMP%203%20Paper-Coming%20Soon-lightgrey?style=for-the-badge&logo=arxiv" alt="Paper"></a>
-        <a href="https://github.com/sanderwood/clamp3"><img src="https://img.shields.io/badge/CLaMP%203%20Code-GitHub-181717?style=for-the-badge&logo=github" alt="GitHub"></a>
-        <a href="https://huggingface.co/sander-wood/clamp3/tree/main"><img src="https://img.shields.io/badge/Model%20Weights-Hugging%20Face-ffcc00?style=for-the-badge&logo=huggingface" alt="Model Weights"></a>
-        <a href="https://huggingface.co/datasets/sander-wood/m4-rag"><img src="https://img.shields.io/badge/M4--RAG%20Pretraining%20Dataset-Hugging%20Face-ffcc00?style=for-the-badge&logo=huggingface" alt="Dataset"></a>
-        <a href="https://huggingface.co/datasets/sander-wood/wikimt-x"><img src="https://img.shields.io/badge/WikiMT--X%20Evaluation%20Benchmark-Hugging%20Face-ffcc00?style=for-the-badge&logo=huggingface" alt="Benchmark"></a>
-    </div>
-    <style>
-        div a {
-            display: inline-block;
-            margin: 5px;
-        }
-        div a img {
-            height: 30px;
-        }
-    </style>
+<div style="text-align: center;">
+    <a href="https://sanderwood.github.io/clamp3/">
+        <img src="https://img.shields.io/badge/CLaMP%203%20Homepage-GitHub-181717?style=for-the-badge&logo=home-assistant" alt="Homepage">
+    </a>
+    <a href="https://arxiv.org/abs/2502.10362">
+        <img src="https://img.shields.io/badge/CLaMP%203%20Paper-Arxiv-red?style=for-the-badge&logo=arxiv" alt="Paper">
+    </a>
+    <a href="https://github.com/sanderwood/clamp3">
+        <img src="https://img.shields.io/badge/CLaMP%203%20Code-GitHub-181717?style=for-the-badge&logo=github" alt="GitHub">
+    </a>
+    <a href="https://huggingface.co/spaces/sander-wood/clamp3">
+        <img src="https://img.shields.io/badge/CLaMP%203%20Demo-Gradio-green?style=for-the-badge&logo=gradio" alt="Demo">
+    </a>
+    <a href="https://huggingface.co/sander-wood/clamp3/tree/main">
+        <img src="https://img.shields.io/badge/Model%20Weights-Hugging%20Face-ffcc00?style=for-the-badge&logo=huggingface" alt="Model Weights">
+    </a>
+    <a href="https://huggingface.co/datasets/sander-wood/m4-rag">
+        <img src="https://img.shields.io/badge/M4--RAG%20Dataset-Hugging%20Face-ffcc00?style=for-the-badge&logo=huggingface" alt="Dataset">
+    </a>
+    <a href="https://huggingface.co/datasets/sander-wood/wikimt-x">
+        <img src="https://img.shields.io/badge/WikiMT--X%20Benchmark-Hugging%20Face-ffcc00?style=for-the-badge&logo=huggingface" alt="Benchmark">
+    </a>
+</div>
+
+<style>
+    div a {
+        display: inline-block;
+        margin: 5px;
+    }
+    div a img {
+        height: 30px;
+    }
+</style>
 """
 
-description = """CLaMP 3 is a **multimodal and multilingual** music information retrieval (MIR) framework, supporting **sheet music, audio, and performance signals** in over **100 languages**. Using **contrastive learning**, it aligns these modalities in a shared space for **cross-modal retrieval**.
+description = """CLaMP 3 is a **multimodal and multilingual** music information retrieval (MIR) framework, supporting **sheet music, audio, and performance signals** in **100 languages**. Using **contrastive learning**, it aligns these modalities in a shared space for **cross-modal retrieval**.
 
 ### 🔍 **How This Demo Works**
 - You can **retrieve music using any text input (in any language) or an image** (`.png`, `.jpg`).  
@@ -41,10 +57,13 @@ description = """CLaMP 3 is a **multimodal and multilingual** music information
 ### ⚠️ **Limitations**
 - This demo retrieves music **only from the WikiMT-X benchmark (1,000 pieces)**.
 - These pieces are **mainly from the U.S. and Western Europe (especially the U.S.)** and **mostly from the 20th century**.
-- The retrieval results are **mostly limited to Western 20th-century music**, so you **won’t** find music from **other regions or historical periods**.
-- If you need retrieval for a **different music collection**, deploy **CLaMP 3 on your own dataset**.
+- Thus, retrieval results are **mostly limited to Western 20th-century music**, so you **won’t** find music from **other regions or historical periods**.
+
+🔧 **Need retrieval for a different music collection?** Deploy **[CLaMP 3](https://github.com/sanderwood/clamp3)**   on your own dataset.  
+Generally, the larger and more diverse the reference music dataset, the better the retrieval quality, increasing the likelihood of finding relevant and accurately matched music.
 
-This demo is for **research purposes only**."""
+**Note: This project is for research use only.**
+"""
 
 # Load BLIP image captioning model and processor
 processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
@@ -246,20 +265,27 @@ def search_wrapper(search_mode, text_input, image_input):
     details = show_details(top_candidate)
     return text_to_use, gr.update(choices=choices), *details
 
+# 定义示例数据（示例数据放在组件定义之后也可以正常运行）
+examples = [
+    ["Text", "classic rock, British, 1960s, upbeat", None],
+    ["Text", "A Latin jazz piece with rhythmic percussion and brass", None],
+    ["Text", "big band, major key, swing, brass-heavy, syncopation, baritone vocal", None],
+    ["Text", "Heartfelt and nostalgic, with a bittersweet, melancholic feel", None],
+    ["Text", "Melodía instrumental en re mayor con progresión armónica repetitiva y fluida", None],
+    ["Text", "D大调四四拍的爱尔兰舞曲", None],
+    ["Text", "Ιερή μουσική με πνευματική ατμόσφαιρα", None],
+    ["Text", "የፍቅር ሙዚቃ ሞቅ እና ስሜታማ ከሆነ ነገር ግን ድንቅ እና አስደሳች ቃላት ያካትታል", None],
+    ["Image", None, "V4EauuhVEw4.jpg"],
+    ["Image", None, "Kw-_Ew5bVxs.jpg"],
+    ["Image", None, "BuYf0taXoNw.jpg"],
+    ["Image", None, "4tDYMayp6Dk.jpg"],
+]
+
 with gr.Blocks() as demo:
     gr.Markdown(title)
     gr.HTML(badges)
     gr.Markdown(description)
-    gr.HTML(
-        """
-        <style>
-          .vertical-radio .gradio-radio label {
-              display: block !important;
-              margin-bottom: 5px;
-          }
-        </style>
-        """
-    )
+
     with gr.Row():
         with gr.Column():
             search_mode = gr.Radio(
@@ -270,13 +296,13 @@ with gr.Blocks() as demo:
                 elem_classes=["vertical-radio"]
             )
             text_input = gr.Textbox(
-                                    placeholder="Describe the music you're looking for (in any language)",
-                                    lines=4
-                                )
+                placeholder="Describe the music you're looking for (in any language)",
+                lines=4
+            )
             image_input = gr.Image(
-                                    label="Or upload an image (PNG, JPG)",
-                                    type="pil"
-                                )
+                label="Or upload an image (PNG, JPG)",
+                type="pil"
+            )
             search_button = gr.Button("Search")
             candidate_radio = gr.Radio(choices=[], label="Select Retrieval Result", interactive=True, elem_classes=["vertical-radio"])
         with gr.Column():
@@ -290,11 +316,32 @@ with gr.Blocks() as demo:
             analysis_box = gr.Textbox(label="Analysis", interactive=False)
             description_box = gr.Textbox(label="Description", interactive=False)
             scene_box = gr.Textbox(label="Scene", interactive=False)
+
+    gr.HTML(
+        """
+        <style>
+          .vertical-radio .gradio-radio label {
+              display: block !important;
+              margin-bottom: 5px;
+          }
+        </style>
+        """
+    )
+
+    gr.Examples(
+        examples=examples,
+        inputs=[search_mode, text_input, image_input],
+        outputs=[text_input, candidate_radio, title_box, artists_box, genre_box, background_box, analysis_box, description_box, scene_box, youtube_box],
+        fn=search_wrapper,
+        cache_examples=False,
+    )
+
     search_button.click(
         fn=search_wrapper,
         inputs=[search_mode, text_input, image_input],
         outputs=[text_input, candidate_radio, title_box, artists_box, genre_box, background_box, analysis_box, description_box, scene_box, youtube_box]
     )
+
     candidate_radio.change(
         fn=show_details,
         inputs=candidate_radio,