File size: 4,458 Bytes
c500600 | 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 | import streamlit as st
import pandas as pd
import pytesseract
from PIL import Image
from rapidfuzz import fuzz, utils
import io
# --- CONFIGURATION ---
st.set_page_config(page_title="AI Student Grader", layout="wide")
st.title("📝 AI Student Answer Grader")
st.markdown("Upload answer sheets and an answer key to automatically calculate marks.")
# --- SIDEBAR SETTINGS ---
st.sidebar.header("Grading Settings")
accuracy_threshold = st.sidebar.slider("Minimum Accuracy Threshold (%)", 0, 100, 70)
marks_per_question = st.sidebar.number_input("Marks per correct answer", value=1.0)
# --- HELPER FUNCTIONS ---
def perform_ocr(image):
"""Extracts text from an uploaded image."""
img = Image.open(image)
# Optional: Add image preprocessing here (grayscale, thresholding)
text = pytesseract.image_to_string(img)
return text.strip()
def compare_answers(student_text, answer_key, threshold):
"""
Compares student text with answer key using Fuzzy Matching.
DeepSeek-R1 style logic: We look for the presence of key concepts.
"""
# Simple line-by-line comparison (assuming 1 question per line or similar structure)
# For complex papers, you'd split by question numbers
score = 0
key_lines = [line.strip() for line in answer_key.split('\n') if line.strip()]
student_lines = [line.strip() for line in student_text.split('\n') if line.strip()]
details = []
for i, correct_ans in enumerate(key_lines):
match_found = False
highest_match = 0
# Compare against each line in student text to find the best match for this answer
for s_line in student_lines:
similarity = fuzz.token_set_ratio(correct_ans, s_line)
if similarity > highest_match:
highest_match = similarity
if highest_match >= threshold:
score += marks_per_question
match_found = True
details.append({
"Question": i + 1,
"Match %": round(highest_match, 2),
"Status": "Correct" if match_found else "Incorrect"
})
return score, details
# --- UI LAYOUT ---
col1, col2 = st.columns(2)
with col1:
st.subheader("1. Reference Answer Key")
key_input_type = st.radio("Key Format", ["Text Input", "Upload Image"])
if key_input_type == "Text Input":
answer_key_text = st.text_area("Paste the correct answers (one per line):")
else:
key_img = st.file_uploader("Upload Answer Key Image", type=['png', 'jpg', 'jpeg'])
if key_img:
answer_key_text = perform_ocr(key_img)
st.text_area("Extracted Key (Edit if needed):", value=answer_key_text)
with col2:
st.subheader("2. Student Answer Sheets")
student_images = st.file_uploader("Upload Student Images (Max 5)", type=['png', 'jpg', 'jpeg'], accept_multiple_files=True)
# --- PROCESSING ---
if st.button("Calculate Marks"):
if not answer_key_text or not student_images:
st.error("Please provide both the answer key and student images.")
else:
results = []
progress_bar = st.progress(0)
for idx, img_file in enumerate(student_images):
# 1. OCR
extracted_text = perform_ocr(img_file)
# 2. Compare
score, details = compare_answers(extracted_text, answer_key_text, accuracy_threshold)
# 3. Store Results
results.append({
"Student Name": img_file.name.split('.')[0], # Uses filename as name
"Raw Score": score,
"Final Marks": f"{score}/{len(answer_key_text.splitlines()) * marks_per_question}",
"Match Percentage": f"{accuracy_threshold}%"
})
progress_bar.progress((idx + 1) / len(student_images))
# --- DISPLAY RESULTS ---
df = pd.DataFrame(results)
st.subheader("📊 Results Overview")
st.table(df)
# --- EXCEL EXPORT ---
output = io.BytesIO()
with pd.ExcelWriter(output, engine='openpyxl') as writer:
df.to_excel(writer, index=False, sheet_name='Grades')
st.download_button(
label="📥 Download Excel Sheet",
data=output.getvalue(),
file_name="student_grades.xlsx",
mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
) |