File size: 13,814 Bytes
e31feff
 
9b15176
2e7b12e
e31feff
 
 
 
 
 
 
 
 
 
 
 
 
9b15176
 
 
 
 
 
 
 
 
 
 
 
25a67ca
2e7b12e
 
 
25a67ca
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2e7b12e
 
 
 
 
 
 
 
 
 
 
 
 
 
25a67ca
2e7b12e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25a67ca
 
2e7b12e
 
25a67ca
2e7b12e
 
25a67ca
 
2e7b12e
 
 
25a67ca
 
 
 
 
 
 
 
 
 
 
 
 
 
2e7b12e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
25a67ca
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e31feff
 
9b15176
 
e31feff
9b15176
e31feff
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9b15176
e31feff
 
 
 
 
 
 
 
 
 
 
 
 
3b773d0
 
 
 
 
 
9b15176
 
e31feff
9b15176
e31feff
9b15176
 
e31feff
2e7b12e
25a67ca
2e7b12e
 
e31feff
 
25a67ca
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e31feff
2e7b12e
25a67ca
 
 
 
 
 
 
 
 
 
 
 
 
2e7b12e
 
e31feff
 
 
3b773d0
 
 
e31feff
 
 
 
 
9b15176
 
 
 
 
e31feff
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3b773d0
 
 
 
e31feff
 
 
 
 
 
3b773d0
2e7b12e
3b773d0
 
 
 
 
 
 
 
e31feff
 
3b773d0
 
 
 
 
e31feff
 
3b773d0
 
e31feff
 
 
3b773d0
 
 
 
9b15176
 
3b773d0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e31feff
 
 
 
3b773d0
e31feff
 
 
3b773d0
 
 
 
 
 
 
 
 
 
 
e31feff
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
import os
import gradio as gr
import re
import folium
from fastai.vision.all import *
from groq import Groq
from PIL import Image

# Load the trained model
learn = load_learner('export.pkl')
labels = learn.dls.vocab

# Initialize Groq client
client = Groq(
    api_key=os.environ.get("GROQ_API_KEY"),
)

def clean_bird_name(name):
    """Clean bird name by removing numbers and special characters, and fix formatting"""
    # Remove numbers and dots at the beginning
    cleaned = re.sub(r'^\d+\.', '', name)
    # Replace underscores with spaces
    cleaned = cleaned.replace('_', ' ')
    # Remove any remaining special characters
    cleaned = re.sub(r'[^\w\s]', '', cleaned)
    # Fix spacing
    cleaned = ' '.join(cleaned.split())
    return cleaned

def get_bird_habitat_map(bird_name, check_tanzania=True):
    """Get habitat map locations for the bird using Groq API"""
    clean_name = clean_bird_name(bird_name)
    
    # First check if the bird is endemic to Tanzania
    if check_tanzania:
        tanzania_check_prompt = f"""
        Is the {clean_name} bird native to or commonly found in Tanzania? 
        Answer with ONLY "yes" or "no".
        """
        
        try:
            tanzania_check = client.chat.completions.create(
                messages=[{"role": "user", "content": tanzania_check_prompt}],
                model="llama-3.3-70b-versatile",
            )
            is_in_tanzania = "yes" in tanzania_check.choices[0].message.content.lower()
        except:
            # Default to showing Tanzania if we can't determine
            is_in_tanzania = True
    else:
        is_in_tanzania = True
    
    # Now get the habitat locations
    prompt = f"""
    Provide a JSON array of the main habitat locations for the {clean_name} bird in the world. 
    Return ONLY a JSON array with 3-5 entries, each containing:
    1. "name": Location name
    2. "lat": Latitude (numeric value)
    3. "lon": Longitude (numeric value)
    4. "description": Brief description of why this is a key habitat (2-3 sentences)
    
    Example format:
    [
      {{"name": "Example Location", "lat": 12.34, "lon": 56.78, "description": "Brief description"}},
      ...
    ]
    
    {'' if is_in_tanzania else 'DO NOT include any locations in Tanzania as this bird is not native to or commonly found there.'}
    """
    
    try:
        chat_completion = client.chat.completions.create(
            messages=[
                {
                    "role": "user",
                    "content": prompt,
                }
            ],
            model="llama-3.3-70b-versatile",
        )
        response = chat_completion.choices[0].message.content
        
        # Extract JSON from response (in case there's additional text)
        import json
        import re
        
        # Find JSON pattern in response
        json_match = re.search(r'\[.*\]', response, re.DOTALL)
        if json_match:
            locations = json.loads(json_match.group())
        else:
            # Fallback if JSON extraction fails
            locations = [
                {"name": "Primary habitat region", "lat": 0, "lon": 0, 
                 "description": "Could not retrieve specific habitat information for this bird."}
            ]
        
        return locations, is_in_tanzania
        
    except Exception as e:
        return [{"name": "Error retrieving data", "lat": 0, "lon": 0, 
                "description": "Please try again or check your connection."}], False

def create_habitat_map(habitat_locations):
    """Create a folium map with the habitat locations"""
    # Find center point based on valid coordinates
    valid_coords = [(loc.get("lat", 0), loc.get("lon", 0)) 
                    for loc in habitat_locations 
                    if loc.get("lat", 0) != 0 or loc.get("lon", 0) != 0]
    
    if valid_coords:
        # Calculate the average of the coordinates
        avg_lat = sum(lat for lat, _ in valid_coords) / len(valid_coords)
        avg_lon = sum(lon for _, lon in valid_coords) / len(valid_coords)
        # Create map centered on the average coordinates
        m = folium.Map(location=[avg_lat, avg_lon], zoom_start=3)
    else:
        # Default world map if no valid coordinates
        m = folium.Map(location=[20, 0], zoom_start=2)
    
    # Add markers for each habitat location
    for location in habitat_locations:
        name = location.get("name", "Unknown")
        lat = location.get("lat", 0)
        lon = location.get("lon", 0)
        description = location.get("description", "No description available")
        
        # Skip invalid coordinates
        if lat == 0 and lon == 0:
            continue
            
        # Add marker
        folium.Marker(
            location=[lat, lon],
            popup=folium.Popup(f"<b>{name}</b><br>{description}", max_width=300),
            tooltip=name
        ).add_to(m)
    
    # Save map to HTML
    map_html = m._repr_html_()
    return map_html

def format_bird_info(raw_info):
    """Improve the formatting of bird information"""
    # Add proper line breaks between sections and ensure consistent heading levels
    formatted = raw_info
    
    # Fix heading levels (make all main sections h3)
    formatted = re.sub(r'#+\s+NOT TYPICALLY FOUND IN TANZANIA', 
                        '<div class="alert alert-warning"><strong>⚠️ NOT TYPICALLY FOUND IN TANZANIA</strong></div>', 
                        formatted)
    
    # Replace markdown headings with HTML headings for better control
    formatted = re.sub(r'#+\s+(.*)', r'<h3>\1</h3>', formatted)
    
    # Add paragraph tags for better spacing
    formatted = re.sub(r'\n\*\s+(.*)', r'<p>• \1</p>', formatted)
    formatted = re.sub(r'\n([^<\n].*)', r'<p>\1</p>', formatted)
    
    # Remove any duplicate paragraph tags
    formatted = formatted.replace('<p><p>', '<p>')
    formatted = formatted.replace('</p></p>', '</p>')
    
    return formatted

def get_bird_info(bird_name):
    """Get detailed information about a bird using Groq API"""
    clean_name = clean_bird_name(bird_name)
    
    prompt = f"""
    Provide detailed information about the {clean_name} bird, including:
    1. Physical characteristics and appearance
    2. Habitat and distribution
    3. Diet and behavior
    4. Migration patterns (emphasize if this pattern has changed in recent years due to climate change)
    5. Conservation status
    
    If this bird is not commonly found in Tanzania, explicitly flag that this bird is "NOT TYPICALLY FOUND IN TANZANIA" at the beginning of your response and explain why its presence might be unusual.
    
    Format your response in markdown for better readability.
    """
    
    try:
        chat_completion = client.chat.completions.create(
            messages=[
                {
                    "role": "user",
                    "content": prompt,
                }
            ],
            model="llama-3.3-70b-versatile",
        )
        return chat_completion.choices[0].message.content
    except Exception as e:
        return f"Error fetching information: {str(e)}"

def predict_and_get_info(img):
    """Predict bird species and get detailed information"""
    # Process the image
    img = PILImage.create(img)
    
    # Get prediction
    pred, pred_idx, probs = learn.predict(img)
    
    # Get top 5 predictions (or all if less than 5)
    num_classes = min(5, len(labels))
    top_indices = probs.argsort(descending=True)[:num_classes]
    top_probs = probs[top_indices]
    top_labels = [labels[i] for i in top_indices]
    
    # Format as dictionary with cleaned names for display
    prediction_results = {clean_bird_name(top_labels[i]): float(top_probs[i]) for i in range(num_classes)}
    
    # Get top prediction (original format for info retrieval)
    top_bird = str(pred)
    # Also keep a clean version for display
    clean_top_bird = clean_bird_name(top_bird)
    
    # Get habitat locations and create map
    habitat_locations, is_in_tanzania = get_bird_habitat_map(top_bird)
    habitat_map_html = create_habitat_map(habitat_locations)
    
    # Get detailed information about the top predicted bird
    bird_info = get_bird_info(top_bird)
    formatted_info = format_bird_info(bird_info)
    
    # Create combined info with map at the top and properly formatted information
    custom_css = """
    <style>
    .bird-container {
        font-family: Arial, sans-serif;
        padding: 10px;
    }
    .map-container {
        height: 400px;
        width: 100%;
        border: 1px solid #ddd;
        border-radius: 8px;
        overflow: hidden;
        margin-bottom: 20px;
    }
    .info-container {
        line-height: 1.6;
    }
    .info-container h3 {
        margin-top: 20px;
        margin-bottom: 10px;
        color: #2c3e50;
        border-bottom: 1px solid #eee;
        padding-bottom: 5px;
    }
    .info-container p {
        margin-bottom: 10px;
    }
    .alert {
        padding: 10px;
        margin-bottom: 15px;
        border-radius: 4px;
    }
    .alert-warning {
        background-color: #fcf8e3;
        border: 1px solid #faebcc;
        color: #8a6d3b;
    }
    </style>
    """
    
    combined_info = f"""
    {custom_css}
    <div class="bird-container">
        <h2>Natural Habitat Map for {clean_top_bird}</h2>
        <div class="map-container">
            {habitat_map_html}
        </div>
        
        <div class="info-container">
            <h2>Detailed Information</h2>
            {formatted_info}
        </div>
    </div>
    """
    
    return prediction_results, combined_info, clean_top_bird

def follow_up_question(question, bird_name):
    """Allow researchers to ask follow-up questions about the identified bird"""
    if not question.strip() or not bird_name:
        return "Please identify a bird first and ask a specific question about it."
    
    prompt = f"""
    The researcher is asking about the {bird_name} bird: "{question}"
    
    Provide a detailed, scientific answer focusing on accurate ornithological information. 
    If the question relates to Tanzania or climate change impacts, emphasize those aspects in your response.
    
    IMPORTANT: Do not repeat basic introductory information about the bird that would have already been provided in a general description.
    Do not start your answer with phrases like "Introduction to the {bird_name}" or similar repetitive headers.
    Directly answer the specific question asked.
    
    Format your response in markdown for better readability.
    """
    
    try:
        chat_completion = client.chat.completions.create(
            messages=[
                {
                    "role": "user",
                    "content": prompt,
                }
            ],
            model="llama-3.3-70b-versatile",
        )
        return chat_completion.choices[0].message.content
    except Exception as e:
        return f"Error fetching information: {str(e)}"

# Create the Gradio interface
with gr.Blocks(theme=gr.themes.Soft()) as app:
    gr.Markdown("# Bird Species Identification for Researchers")
    gr.Markdown("Upload an image to identify bird species and get detailed information relevant to research in Tanzania and climate change studies.")
    
    # Store the current bird for context
    current_bird = gr.State("")
    
    # Main identification section
    with gr.Row():
        with gr.Column(scale=1):
            input_image = gr.Image(type="pil", label="Upload Bird Image")
            submit_btn = gr.Button("Identify Bird", variant="primary")
        
        with gr.Column(scale=2):
            prediction_output = gr.Label(label="Top 5 Predictions", num_top_classes=5)
            bird_info_output = gr.HTML(label="Bird Information")
    
    # Clear divider
    gr.Markdown("---")
    
    # Follow-up question section with improved UI
    gr.Markdown("## Research Questions")
    
    conversation_history = gr.Markdown("")
    
    with gr.Row():
        follow_up_input = gr.Textbox(
            label="Ask a question about this bird", 
            placeholder="Example: How has climate change affected this bird's migration pattern?",
            lines=2
        )
    
    with gr.Row():
        follow_up_btn = gr.Button("Submit Question", variant="primary")
        clear_btn = gr.Button("Clear Conversation")
    
    # Set up event handlers
    def process_image(img):
        if img is None:
            return None, "Please upload an image", "", ""
        
        try:
            pred_results, info, clean_bird_name = predict_and_get_info(img)
            return pred_results, info, clean_bird_name, ""
        except Exception as e:
            return None, f"Error processing image: {str(e)}", "", ""
    
    def update_conversation(question, bird_name, history):
        if not question.strip():
            return history
        
        answer = follow_up_question(question, bird_name)
        
        # Format the conversation with clear separation
        new_exchange = f"""
### Question:
{question}

### Answer:
{answer}

---
"""
        updated_history = new_exchange + history
        return updated_history
    
    def clear_conversation_history():
        return ""
    
    submit_btn.click(
        process_image,
        inputs=[input_image],
        outputs=[prediction_output, bird_info_output, current_bird, conversation_history]
    )
    
    follow_up_btn.click(
        update_conversation,
        inputs=[follow_up_input, current_bird, conversation_history],
        outputs=[conversation_history]
    ).then(
        lambda: "", 
        outputs=follow_up_input
    )
    
    clear_btn.click(
        clear_conversation_history,
        outputs=[conversation_history]
    )

# Launch the app
app.launch(share=True)