import os
import gradio as gr
import re
import folium
from fastai.vision.all import *
from groq import Groq
from PIL import Image
import time
import json
from functools import lru_cache
# 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"),
)
# Cache directory for API responses
os.makedirs("cache", exist_ok=True)
# Language translations
translations = {
"en": {
"app_title": "AvianEye Tanzania",
"app_description": "🔍 Upload a bird photo to instantly identify species and access comprehensive data on habitats, behaviors, and climate change impacts. A powerful tool for Tanzania-based ornithological research.",
"upload_label": "Upload Bird Image",
"identify_button": "Identify Bird",
"predictions_label": "Top 5 Predictions",
"bird_info_label": "Bird Information",
"research_questions": "Research Questions",
"question_placeholder": "Example: How has climate change affected this bird's migration pattern?",
"question_label": "Ask a question about this bird",
"submit_question": "Submit Question",
"clear_conversation": "Clear Conversation",
"upload_prompt": "Please upload an image",
"question_title": "Question:",
"answer_title": "Answer:",
"habitat_map_title": "Natural Habitat Map for",
"detailed_info_title": "Detailed Information",
"language_label": "Language / Lugha",
"loading": "Processing...",
"low_confidence": "Low confidence prediction. Results may not be accurate.",
"not_a_bird": "The image may not contain a bird. Please upload a clear image of a bird.",
"other_message": "This bird is not in our trained dataset or the image may not be of a bird. Please try uploading a different image."
},
"sw": {
"app_title": "Mtafiti wa Ndege: Utambuzi wa Kiotomatiki kwa Watafiti",
"app_description": "🔍 Pakia picha ya ndege ili kutambua spishi mara moja na kupata data kamili kuhusu makazi, tabia, na athari za mabadiliko ya tabianchi. Zana yenye nguvu kwa utafiti wa ndege nchini Tanzania.",
"upload_label": "Pakia Picha ya Ndege",
"identify_button": "Tambua Ndege",
"predictions_label": "Utabiri Bora 5",
"bird_info_label": "Taarifa za Ndege",
"research_questions": "Maswali ya Utafiti",
"question_placeholder": "Mfano: Je, mabadiliko ya tabianchi yameathiri vipi mfumo wa uhamiaji wa ndege huyu?",
"question_label": "Uliza swali kuhusu ndege huyu",
"submit_question": "Wasilisha Swali",
"clear_conversation": "Futa Mazungumzo",
"upload_prompt": "Tafadhali pakia picha",
"question_title": "Swali:",
"answer_title": "Jibu:",
"habitat_map_title": "Ramani ya Makazi Asilia ya",
"detailed_info_title": "Taarifa za Kina",
"language_label": "Language / Lugha",
"loading": "Inachakata...",
"low_confidence": "Utabiri wa uhakika mdogo. Matokeo yanaweza kuwa si sahihi.",
"not_a_bird": "Picha inaweza isiwe ya ndege. Tafadhali pakia picha wazi ya ndege.",
"other_message": "Ndege huyu haipatikani katika hifadhidata yetu au picha inaweza isiwe ya ndege. Tafadhali jaribu kupakia picha nyingine."
}
}
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_cache_path(function_name, key):
"""Generate a cache file path"""
safe_key = re.sub(r'[^\w]', '_', key)
return f"cache/{function_name}_{safe_key}.json"
def save_to_cache(function_name, key, data):
"""Save API response to cache"""
try:
cache_path = get_cache_path(function_name, key)
with open(cache_path, 'w') as f:
json.dump({"data": data, "timestamp": time.time()}, f)
except Exception as e:
print(f"Error saving to cache: {e}")
def load_from_cache(function_name, key, max_age=86400): # Default max age: 1 day
"""Load API response from cache if it exists and is not too old"""
try:
cache_path = get_cache_path(function_name, key)
if os.path.exists(cache_path):
with open(cache_path, 'r') as f:
cached = json.load(f)
if time.time() - cached["timestamp"] < max_age:
return cached["data"]
except Exception as e:
print(f"Error loading from cache: {e}")
return None
def is_likely_bird_image(img):
"""Basic check to see if the image might contain a bird"""
try:
# Convert to numpy array for analysis
img_array = np.array(img)
# Simple checks that might indicate a bird isn't present:
# 1. Check if image is too dark or too bright overall
mean_brightness = np.mean(img_array)
if mean_brightness < 20 or mean_brightness > 235:
return False
# 2. Check if image has very little color variation (might be a solid background)
std_dev = np.std(img_array)
if std_dev < 15:
return False
# 3. If image is very small, it might not be a useful bird photo
if img_array.shape[0] < 100 or img_array.shape[1] < 100:
return False
return True
except:
# If any error occurs during the check, assume it might be a bird
return True
def get_bird_habitat_map(bird_name, check_tanzania=True):
"""Get habitat map locations for the bird using Groq API with caching"""
clean_name = clean_bird_name(bird_name)
# Check cache for Tanzania check
tanzania_cache_key = f"{clean_name}_tanzania"
cached_tanzania = load_from_cache("tanzania_check", tanzania_cache_key)
if cached_tanzania is not None:
is_in_tanzania = cached_tanzania
else:
# 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="deepseek-r1-distill-llama-70b",
)
is_in_tanzania = "yes" in tanzania_check.choices[0].message.content.lower()
# Cache result
save_to_cache("tanzania_check", tanzania_cache_key, is_in_tanzania)
except:
# Default to showing Tanzania if we can't determine
is_in_tanzania = True
else:
is_in_tanzania = True
# Check cache for habitat locations
habitat_cache_key = f"{clean_name}_habitat"
cached_habitat = load_from_cache("habitat", habitat_cache_key)
if cached_habitat is not None:
return cached_habitat, is_in_tanzania
# 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="deepseek-r1-distill-llama-70b",
)
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."}
]
# Cache the result
save_to_cache("habitat", habitat_cache_key, locations)
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"{name}
{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, language="en"):
"""Improve the formatting of bird information"""
# Add proper line breaks between sections and ensure consistent heading levels
formatted = raw_info
# Get translation of warning text based on language
warning_text = "NOT TYPICALLY FOUND IN TANZANIA"
warning_translation = "HAPATIKANI SANA TANZANIA" if language == "sw" else warning_text
# Fix heading levels (make all main sections h3)
formatted = re.sub(r'#+\s+' + warning_text,
f'⚠️ {warning_translation}
',
formatted)
# Replace markdown headings with HTML headings for better control
formatted = re.sub(r'#+\s+(.*)', r'\1
', formatted)
# Add paragraph tags for better spacing
formatted = re.sub(r'\n\*\s+(.*)', r'• \1
', formatted)
formatted = re.sub(r'\n([^<\n].*)', r'\1
', formatted)
# Remove any duplicate paragraph tags
formatted = formatted.replace('', '
')
formatted = formatted.replace('
', '')
return formatted
def get_bird_info(bird_name, language="en"):
"""Get detailed information about a bird using Groq API with caching"""
clean_name = clean_bird_name(bird_name)
# Check cache first
cache_key = f"{clean_name}_{language}"
cached_info = load_from_cache("bird_info", cache_key)
if cached_info is not None:
return cached_info
# Adjust language for the prompt
lang_instruction = ""
if language == "sw":
lang_instruction = " Provide your response in Swahili language."
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.{lang_instruction}
"""
try:
chat_completion = client.chat.completions.create(
messages=[
{
"role": "user",
"content": prompt,
}
],
model="deepseek-r1-distill-llama-70b",
)
response = chat_completion.choices[0].message.content
# Cache the result
save_to_cache("bird_info", cache_key, response)
return response
except Exception as e:
error_msg = "Hitilafu katika kupata taarifa" if language == "sw" else "Error fetching information"
return f"{error_msg}: {str(e)}"
def create_message_html(message, icon="🔍", language="en"):
"""Create a styled message container for notifications"""
custom_css = """
"""
html = f"""
{custom_css}
"""
return html
def predict_and_get_info(img, language="en"):
"""Predict bird species and get detailed information"""
# Get translations
t = translations[language]
# Check if an image was provided
if img is None:
message = t['upload_prompt']
return None, create_message_html(message, "📷", language), "", ""
# Basic check if the image might contain a bird
if not is_likely_bird_image(img):
message = t['not_a_bird']
return None, create_message_html(message, "⚠️", language), "", ""
try:
# 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)
# Check if the model's confidence is low
if float(top_probs[0]) < 0.4:
low_confidence_warning = t['low_confidence']
else:
low_confidence_warning = ""
# Check if the top prediction is "Other" and has high confidence
if "other" in clean_top_bird.lower():
# Create a message informing the user that the bird wasn't recognized
other_message = t['other_message']
combined_info = create_message_html(other_message, "🔍", language)
return prediction_results, combined_info, clean_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, language)
formatted_info = format_bird_info(bird_info, language)
# Create combined info with map at the top and properly formatted information
custom_css = """
"""
# Add low confidence warning if needed
confidence_warning_html = f'{low_confidence_warning}
' if low_confidence_warning else ''
combined_info = f"""
{custom_css}
{confidence_warning_html}
{t['habitat_map_title']} {clean_top_bird}
{habitat_map_html}
{t['detailed_info_title']}
{formatted_info}