feat: app update with UI improvements and example texts

app.py

@@ -231,12 +231,41 @@ def predict_language(audio_path, fam_k=1, fam_thresh=0.0, super_k=1, super_thres
         print(f"Error during inference: {e}")
         raise gr.Error(f"Processing failed: {str(e)}")
 
+import base64
+import re
+
+# --- Load icon.png as base64 ---
+try:
+    with open("icon.png", "rb") as image_file:
+        encoded_string = base64.b64encode(image_file.read()).decode("utf-8")
+        icon_html = f'<img src="data:image/png;base64,{encoded_string}" style="height: 80px; margin-bottom: 10px;" alt="Tachiwin Icon" />'
+except Exception as e:
+    icon_html = ""
+
+# --- Load curated examples ---
+curated_examples = []
+try:
+    with open("results.txt", "r", encoding="utf-8") as f:
+        results_content = f.read()
+    
+    # Split by the sample header
+    sample_blocks = re.split(r'📊 Sample (\d+)', results_content)
+    for i in range(1, len(sample_blocks), 2):
+        sample_num = sample_blocks[i]
+        block = sample_blocks[i+1].strip()
+        audio_path = f"samples/sample{sample_num}.wav"
+        if os.path.exists(audio_path):
+            curated_examples.append([audio_path, f"Sample {sample_num}:\n{block}"])
+except Exception as e:
+    print(f"Warning: Could not parse results.txt: {e}")
+
 # === UI COMPONENTS ===
 with gr.Blocks(theme=gr.themes.Ocean()) as demo:
     gr.HTML(
-        """
+        f"""
         <div style="text-align: center; padding: 30px; background: linear-gradient(120deg, rgb(2, 132, 199) 0%, rgb(16, 185, 129) 60%, rgb(5, 150, 105) 100%); color: white; border-radius: 15px; margin-bottom: 25px; box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);">
-            <h1 style="color: white; margin: 0; font-size: 2.5em;">🦡 Tachiwin Language Identifier 🦡</h1>
+            {icon_html}
+            <h1 style="color: white; margin: 0; font-size: 2.5em;">🦡 Tachiwin AudioId 🦡</h1>
             <p style="font-size: 1.2em; opacity: 0.9; margin-top: 10px;">Identify any of the 68 languages of Mexico and their 360 variants</p>
         </div>
         """
@@ -292,16 +321,18 @@ with gr.Blocks(theme=gr.themes.Ocean()) as demo:
     gr.Markdown(
         """
         ---
-        ### ℹ️ About this Model
-        Tachiwin Multilingual Language Classifier is a finetune/fork or encoded-only whisper architecture trained to recognize any of the 68 indigenous superlanguages of México and their 360 variants.
-        **Accuracy Overview:**
-        - **Language Family**: ~73% 
-        - **Superlanguage**: ~59% 
-        - **Language Code**: ~52%
-        
-        *Developed by Tachiwin. May the indigenous languages never be lost.*
+        ### 📋 Curated Samples for Reference
+        Select any of the sample audios below to load them for inference. The expected results (Truth) and model predictions from the finetune phase are shown below for reference.
         """
     )
+    
+    with gr.Row():
+        example_text = gr.Textbox(label="Sample Ground Truth & Finetune Prediction", interactive=False, lines=6)
+
+    gr.Examples(
+        examples=curated_examples,
+        inputs=[audio_input, example_text],
+    )
 
 if __name__ == "__main__":
     demo.launch(ssr_mode=False)

results.txt

@@ -0,0 +1,176 @@
+ 📊 Sample 0
+Truth: Mayense → Ch'ol → ctu
+Pred:  Mayense → Ch'ol → ctu
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 60
+Truth: Mayense → Tzeltal → tzh
+Pred:  Mayense → Maya → yua
+Match: 0.33 ✅
+  Family:  ✅
+  Super:   ❌
+  Code:    ❌
+
+
+📊 Sample 200
+Truth: Yuto-Nahua → Mayo → mfy
+Pred:  Yuto-Nahua → Mayo → mfy
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 220
+Truth: Yuto-Nahua → Náhuatl → nlv
+Pred:  Yuto-Nahua → Náhuatl → ngu
+Match: 0.67 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ❌
+
+📊 Sample 240
+Truth: Mayense → Maya → yua
+Pred:  Mayense → Maya → yua
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 260
+Truth: Yuto-Nahua → Yaqui → yaq
+Pred:  Yuto-Nahua → Yaqui → yaq
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 280
+Truth: Otomangue → Chatino → ctp
+Pred:  Otomangue → Chatino → ctp
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 400
+Truth: Otomangue → Zapoteco → zaa
+Pred:  Otomangue → Zapoteco → zaa
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 420
+Truth: Otomangue → Mazahua → mmc
+Pred:  Otomangue → Mazahua → mmc
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 440
+Truth: Yuto-Nahua → Cora → crn
+Pred:  Yuto-Nahua → Cora → crn
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 480
+Truth: Otomangue → Mixteco → mix
+Pred:  Otomangue → Mixteco → mix
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 600
+Truth: Yuto-Nahua → Guarijío → var
+Pred:  Yuto-Nahua → Guarijío → var
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 660
+Truth: Yuto-Nahua → Mayo → mfy
+Pred:  Yuto-Nahua → Mayo → mfy
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 820
+Truth: Otomangue → Chatino → ctp
+Pred:  Otomangue → Chatino → ctp
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 1020
+Truth: Mayense → Tzeltal → tzh
+Pred:  Mayense → Tzeltal → tzh
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 1080
+Truth: Yuto-Nahua → Huichol → hch
+Pred:  Yuto-Nahua → Huichol → hch
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 1100
+Truth: Otomangue → Triqui → trs
+Pred:  Otomangue → Triqui → trs
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 1180
+Truth: Otomangue → Otomí → ott
+Pred:  Otomangue → Otomí → ott
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 1200
+Truth: Otomangue → Zapoteco → zaa
+Pred:  Otomangue → Zapoteco → zaa
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 1480
+Truth: Mayense → Tojolabal → toj
+Pred:  Mayense → Tojolabal → tzh
+Match: 0.67 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ❌
+
+📊 Sample 1580
+Truth: Purépecha → Purépecha → tsz
+Pred:  Purépecha → Purépecha → tsz
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅
+
+📊 Sample 1880
+Truth: Yuto-Nahua → Tepehuano → ntp
+Pred:  Yuto-Nahua → Tepehuano → ntp
+Match: 1.00 ✅
+  Family:  ✅
+  Super:   ✅
+  Code:    ✅