Spaces:
Running
Running
Update check.py
Browse files
check.py
CHANGED
@@ -1,3 +1,5 @@
|
|
|
|
|
|
1 |
import os
|
2 |
import tempfile
|
3 |
import json
|
@@ -5,256 +7,170 @@ import numpy as np
|
|
5 |
import cv2
|
6 |
from PIL import Image
|
7 |
from pdf2image import convert_from_bytes
|
8 |
-
from fastapi import
|
9 |
from fastapi.responses import JSONResponse, StreamingResponse
|
10 |
-
import
|
11 |
-
from fastapi import APIRouter, HTTPException, Path
|
12 |
|
|
|
13 |
|
14 |
-
#
|
15 |
GENAI_API_KEY = os.getenv("GENAI_API_KEY")
|
16 |
if not GENAI_API_KEY:
|
17 |
raise Exception("GENAI_API_KEY not set in environment")
|
18 |
-
|
19 |
-
# Import the Google GenAI client libraries.
|
20 |
-
from google import genai
|
21 |
-
from google.genai import types
|
22 |
-
|
23 |
-
# Initialize the GenAI client with the API key.
|
24 |
client = genai.Client(api_key=GENAI_API_KEY)
|
25 |
|
26 |
-
|
27 |
-
|
28 |
-
# Use system temporary directory to store the results file.
|
29 |
TEMP_FOLDER = tempfile.gettempdir()
|
30 |
RESULT_FILE = os.path.join(TEMP_FOLDER, "result_cards.json")
|
31 |
|
32 |
|
33 |
-
##############################################################
|
34 |
-
# Preprocessing & Extraction Functions
|
35 |
-
##############################################################
|
36 |
-
|
37 |
def extract_json_from_output(output_str: str):
|
38 |
-
""
|
39 |
-
|
40 |
-
"""
|
41 |
-
start = output_str.find('{')
|
42 |
-
end = output_str.rfind('}')
|
43 |
if start == -1 or end == -1:
|
44 |
-
print("No JSON block found in the output.")
|
45 |
return None
|
46 |
-
json_str = output_str[start:end+1]
|
47 |
try:
|
48 |
-
return json.loads(
|
49 |
-
except json.JSONDecodeError
|
50 |
-
print("Error decoding JSON:", e)
|
51 |
return None
|
52 |
|
53 |
|
54 |
def parse_all_answers(image_input: Image.Image) -> str:
|
55 |
-
"""
|
56 |
-
Extracts answers from an image of a 15-question answer sheet.
|
57 |
-
Returns the raw JSON string response from the model.
|
58 |
-
"""
|
59 |
output_format = """
|
60 |
Answer in the following JSON format. Do not write anything else:
|
61 |
-
{
|
62 |
-
"Answers": {
|
63 |
-
"1": "<option or text>",
|
64 |
-
"2": "<option or text>",
|
65 |
-
"3": "<option or text>",
|
66 |
-
"4": "<option or text>",
|
67 |
-
"5": "<option or text>",
|
68 |
-
"6": "<option or text>",
|
69 |
-
"7": "<option or text>",
|
70 |
-
"8": "<option or text>",
|
71 |
-
"9": "<option or text>",
|
72 |
-
"10": "<option or text>",
|
73 |
-
"11": "<free-text answer>",
|
74 |
-
"12": "<free-text answer>",
|
75 |
-
"13": "<free-text answer>",
|
76 |
-
"14": "<free-text answer>",
|
77 |
-
"15": "<free-text answer>"
|
78 |
-
}
|
79 |
-
}
|
80 |
"""
|
81 |
prompt = f"""
|
82 |
-
You are an assistant that extracts answers from an image.
|
83 |
-
|
84 |
-
For questions 1 to 10, the answers are multiple-choice selections.
|
85 |
-
For questions 11 to 15, the answers are free-text responses.
|
86 |
-
Extract the answer for each question (1 to 15) and provide the result in JSON using the format below:
|
87 |
{output_format}
|
88 |
"""
|
89 |
response = client.models.generate_content(
|
90 |
-
model="gemini-2.0-flash",
|
91 |
-
contents=[prompt, image_input]
|
92 |
)
|
93 |
return response.text
|
94 |
|
95 |
|
96 |
def parse_info(image_input: Image.Image) -> str:
|
97 |
-
"""
|
98 |
-
Extracts candidate information including name, number, country, level and paper from an image.
|
99 |
-
Returns the raw JSON string response from the model.
|
100 |
-
"""
|
101 |
output_format = """
|
102 |
Answer in the following JSON format. Do not write anything else:
|
103 |
-
{
|
104 |
-
"Candidate Info": {
|
105 |
-
"Name": "<name>",
|
106 |
-
"Number": "<number>",
|
107 |
-
"Country": "<country>",
|
108 |
-
"Level": "<level>",
|
109 |
-
"Paper": "<paper>"
|
110 |
-
}
|
111 |
-
}
|
112 |
"""
|
113 |
prompt = f"""
|
114 |
-
You are an assistant that extracts candidate
|
115 |
-
|
116 |
-
Extract the information accurately and provide the result in JSON using the following format:
|
117 |
{output_format}
|
118 |
"""
|
119 |
response = client.models.generate_content(
|
120 |
-
model="gemini-2.0-flash",
|
121 |
-
contents=[prompt, image_input]
|
122 |
)
|
123 |
return response.text
|
124 |
|
125 |
|
126 |
def calculate_result(student_answers: dict, correct_answers: dict) -> dict:
|
127 |
-
""
|
128 |
-
|
129 |
-
|
130 |
-
"""
|
131 |
-
student_all = student_answers.get("Answers", {})
|
132 |
-
correct_all = correct_answers.get("Answers", {})
|
133 |
-
total_questions = 15
|
134 |
marks = 0
|
135 |
detailed = {}
|
136 |
-
|
137 |
-
|
138 |
-
|
139 |
-
|
140 |
-
if
|
|
|
141 |
marks += 1
|
142 |
-
|
143 |
-
else:
|
144 |
-
detailed[q] = {"Student": stud_ans, "Correct": corr_ans, "Result": "Incorrect"}
|
145 |
-
|
146 |
-
percentage = (marks / total_questions) * 100
|
147 |
-
return {
|
148 |
-
"Total Marks": marks,
|
149 |
-
"Total Questions": total_questions,
|
150 |
-
"Percentage": percentage,
|
151 |
-
"Detailed Results": detailed
|
152 |
-
}
|
153 |
|
154 |
|
155 |
def load_answer_key(pdf_bytes: bytes) -> dict:
|
156 |
-
"""
|
157 |
-
Converts a PDF (as bytes) to images, takes the last page, and parses the answers.
|
158 |
-
Returns the parsed JSON answer key.
|
159 |
-
"""
|
160 |
images = convert_from_bytes(pdf_bytes)
|
161 |
-
|
162 |
-
|
163 |
-
return extract_json_from_output(
|
164 |
-
|
165 |
|
166 |
-
##############################################################
|
167 |
-
# FastAPI Endpoints
|
168 |
-
##############################################################
|
169 |
|
170 |
-
@router.post("/process")
|
171 |
async def process_pdfs(
|
172 |
-
|
173 |
-
paper_k_pdf:
|
174 |
):
|
175 |
try:
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
-
|
183 |
-
|
184 |
-
|
185 |
-
# Convert the student answer PDF to images (each page = one student)
|
186 |
-
student_images = convert_from_bytes(student_pdf_bytes)
|
187 |
all_results = []
|
188 |
-
|
189 |
-
for idx, page in enumerate(
|
190 |
-
#
|
191 |
-
|
192 |
-
h, w =
|
193 |
mask = np.zeros((h, w), dtype="uint8")
|
194 |
top, bottom = int(h * 0.10), int(h * 0.75)
|
195 |
cv2.rectangle(mask, (0, top), (w, h - bottom), 255, -1)
|
196 |
-
|
197 |
coords = cv2.findNonZero(mask)
|
198 |
if coords is None:
|
199 |
continue
|
200 |
x, y, mw, mh = cv2.boundingRect(coords)
|
201 |
-
cand_img = Image.fromarray(cv2.cvtColor(
|
202 |
-
|
203 |
-
#
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
#
|
208 |
-
|
209 |
-
stud_answers = extract_json_from_output(
|
210 |
-
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
-
|
220 |
-
|
221 |
-
|
222 |
-
|
|
|
|
|
|
|
|
|
223 |
with open(RESULT_FILE, "w", encoding="utf-8") as f:
|
224 |
json.dump({"results": all_results}, f, indent=2)
|
225 |
-
|
226 |
return JSONResponse(content={"results": all_results})
|
227 |
-
|
|
|
|
|
228 |
except Exception as e:
|
229 |
-
raise HTTPException(
|
230 |
|
231 |
|
232 |
-
@router.get("/download")
|
233 |
async def download_results():
|
234 |
-
"""
|
235 |
-
Returns the result JSON file stored in the temporary folder.
|
236 |
-
"""
|
237 |
if not os.path.exists(RESULT_FILE):
|
238 |
-
raise HTTPException(
|
239 |
return StreamingResponse(
|
240 |
open(RESULT_FILE, "rb"),
|
241 |
media_type="application/json",
|
242 |
-
headers={"Content-Disposition": "attachment; filename=result_cards.json"}
|
243 |
)
|
244 |
|
245 |
|
246 |
-
@router.get("/")
|
247 |
-
async def
|
248 |
-
return {
|
249 |
-
"message": "Welcome to the Student Result Card API (Paper K only).",
|
250 |
-
"usage": (
|
251 |
-
"POST two PDFs to /process: "
|
252 |
-
"(1) original answer sheet PDF, "
|
253 |
-
"(2) Paper K answer-key PDF. "
|
254 |
-
"Then GET /download to retrieve the graded results."
|
255 |
-
)
|
256 |
-
}
|
257 |
|
258 |
|
259 |
-
|
260 |
-
|
|
|
|
1 |
+
# check.py
|
2 |
+
|
3 |
import os
|
4 |
import tempfile
|
5 |
import json
|
|
|
7 |
import cv2
|
8 |
from PIL import Image
|
9 |
from pdf2image import convert_from_bytes
|
10 |
+
from fastapi import APIRouter, UploadFile, File, HTTPException
|
11 |
from fastapi.responses import JSONResponse, StreamingResponse
|
12 |
+
from google import genai
|
|
|
13 |
|
14 |
+
router = APIRouter(prefix="/check", tags=["check"])
|
15 |
|
16 |
+
# GenAI client
|
17 |
GENAI_API_KEY = os.getenv("GENAI_API_KEY")
|
18 |
if not GENAI_API_KEY:
|
19 |
raise Exception("GENAI_API_KEY not set in environment")
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
client = genai.Client(api_key=GENAI_API_KEY)
|
21 |
|
22 |
+
# Temp storage for results
|
|
|
|
|
23 |
TEMP_FOLDER = tempfile.gettempdir()
|
24 |
RESULT_FILE = os.path.join(TEMP_FOLDER, "result_cards.json")
|
25 |
|
26 |
|
|
|
|
|
|
|
|
|
27 |
def extract_json_from_output(output_str: str):
|
28 |
+
start = output_str.find("{")
|
29 |
+
end = output_str.rfind("}")
|
|
|
|
|
|
|
30 |
if start == -1 or end == -1:
|
|
|
31 |
return None
|
|
|
32 |
try:
|
33 |
+
return json.loads(output_str[start : end + 1])
|
34 |
+
except json.JSONDecodeError:
|
|
|
35 |
return None
|
36 |
|
37 |
|
38 |
def parse_all_answers(image_input: Image.Image) -> str:
|
|
|
|
|
|
|
|
|
39 |
output_format = """
|
40 |
Answer in the following JSON format. Do not write anything else:
|
41 |
+
{ "Answers": { "1": "<…>", …, "15": "<…>" } }
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
"""
|
43 |
prompt = f"""
|
44 |
+
You are an assistant that extracts answers from an image of a 15-question sheet.
|
45 |
+
Provide ONLY JSON in this format:
|
|
|
|
|
|
|
46 |
{output_format}
|
47 |
"""
|
48 |
response = client.models.generate_content(
|
49 |
+
model="gemini-2.0-flash", contents=[prompt, image_input]
|
|
|
50 |
)
|
51 |
return response.text
|
52 |
|
53 |
|
54 |
def parse_info(image_input: Image.Image) -> str:
|
|
|
|
|
|
|
|
|
55 |
output_format = """
|
56 |
Answer in the following JSON format. Do not write anything else:
|
57 |
+
{ "Candidate Info": { "Name": "<…>", "Number": "<…>", "Country": "<…>", "Level": "<…>", "Paper": "<…>" } }
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
"""
|
59 |
prompt = f"""
|
60 |
+
You are an assistant that extracts candidate info from an image.
|
61 |
+
Provide ONLY JSON in this format:
|
|
|
62 |
{output_format}
|
63 |
"""
|
64 |
response = client.models.generate_content(
|
65 |
+
model="gemini-2.0-flash", contents=[prompt, image_input]
|
|
|
66 |
)
|
67 |
return response.text
|
68 |
|
69 |
|
70 |
def calculate_result(student_answers: dict, correct_answers: dict) -> dict:
|
71 |
+
student_all = (student_answers or {}).get("Answers", {})
|
72 |
+
correct_all = (correct_answers or {}).get("Answers", {})
|
73 |
+
total = 15
|
|
|
|
|
|
|
|
|
74 |
marks = 0
|
75 |
detailed = {}
|
76 |
+
for q in map(str, range(1, total + 1)):
|
77 |
+
stud = (student_all.get(q) or "").strip()
|
78 |
+
corr = (correct_all.get(q) or "").strip()
|
79 |
+
ok = stud == corr
|
80 |
+
detailed[q] = {"Student": stud, "Correct": corr, "Result": "Correct" if ok else "Incorrect"}
|
81 |
+
if ok:
|
82 |
marks += 1
|
83 |
+
return {"Total Marks": marks, "Total Questions": total, "Percentage": marks / total * 100, "Detailed Results": detailed}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
84 |
|
85 |
|
86 |
def load_answer_key(pdf_bytes: bytes) -> dict:
|
|
|
|
|
|
|
|
|
87 |
images = convert_from_bytes(pdf_bytes)
|
88 |
+
last_page = images[-1]
|
89 |
+
resp = parse_all_answers(last_page)
|
90 |
+
return extract_json_from_output(resp)
|
|
|
91 |
|
|
|
|
|
|
|
92 |
|
93 |
+
@router.post("/process", summary="Grade student sheets (Paper K only)")
|
94 |
async def process_pdfs(
|
95 |
+
student_pdf: UploadFile = File(..., description="Student sheets PDF"),
|
96 |
+
paper_k_pdf: UploadFile = File(..., description="Answer key PDF for Paper K"),
|
97 |
):
|
98 |
try:
|
99 |
+
stud_bytes = await student_pdf.read()
|
100 |
+
key_bytes = await paper_k_pdf.read()
|
101 |
+
|
102 |
+
answer_key = load_answer_key(key_bytes)
|
103 |
+
if answer_key is None:
|
104 |
+
raise HTTPException(400, detail="Could not parse Paper K answer key.")
|
105 |
+
|
106 |
+
student_pages = convert_from_bytes(stud_bytes)
|
|
|
|
|
|
|
107 |
all_results = []
|
108 |
+
|
109 |
+
for idx, page in enumerate(student_pages, start=1):
|
110 |
+
# crop candidate-info
|
111 |
+
cv = cv2.cvtColor(np.array(page), cv2.COLOR_RGB2BGR)
|
112 |
+
h, w = cv.shape[:2]
|
113 |
mask = np.zeros((h, w), dtype="uint8")
|
114 |
top, bottom = int(h * 0.10), int(h * 0.75)
|
115 |
cv2.rectangle(mask, (0, top), (w, h - bottom), 255, -1)
|
116 |
+
crop = cv2.bitwise_and(cv, cv, mask=mask)
|
117 |
coords = cv2.findNonZero(mask)
|
118 |
if coords is None:
|
119 |
continue
|
120 |
x, y, mw, mh = cv2.boundingRect(coords)
|
121 |
+
cand_img = Image.fromarray(cv2.cvtColor(crop[y : y + mh, x : x + mw], cv2.COLOR_BGR2RGB))
|
122 |
+
|
123 |
+
# parse candidate info
|
124 |
+
info_txt = parse_info(cand_img)
|
125 |
+
candidate_info = extract_json_from_output(info_txt) or {}
|
126 |
+
|
127 |
+
# parse student answers
|
128 |
+
stud_txt = parse_all_answers(page)
|
129 |
+
stud_answers = extract_json_from_output(stud_txt)
|
130 |
+
if stud_answers is None:
|
131 |
+
raise HTTPException(400, detail=f"Failed to parse answers on page {idx}.")
|
132 |
+
|
133 |
+
# grade
|
134 |
+
result = calculate_result(stud_answers, answer_key)
|
135 |
+
|
136 |
+
all_results.append(
|
137 |
+
{
|
138 |
+
"Student Index": idx,
|
139 |
+
"Candidate Info": candidate_info.get("Candidate Info", {}),
|
140 |
+
"Student Answers": stud_answers,
|
141 |
+
"Correct Answer Key": answer_key,
|
142 |
+
"Result": result,
|
143 |
+
}
|
144 |
+
)
|
145 |
+
|
146 |
+
# write file
|
147 |
with open(RESULT_FILE, "w", encoding="utf-8") as f:
|
148 |
json.dump({"results": all_results}, f, indent=2)
|
149 |
+
|
150 |
return JSONResponse(content={"results": all_results})
|
151 |
+
|
152 |
+
except HTTPException:
|
153 |
+
raise
|
154 |
except Exception as e:
|
155 |
+
raise HTTPException(500, detail=str(e))
|
156 |
|
157 |
|
158 |
+
@router.get("/download", summary="Download latest grading results")
|
159 |
async def download_results():
|
|
|
|
|
|
|
160 |
if not os.path.exists(RESULT_FILE):
|
161 |
+
raise HTTPException(404, detail="No results available. Run /check/process first.")
|
162 |
return StreamingResponse(
|
163 |
open(RESULT_FILE, "rb"),
|
164 |
media_type="application/json",
|
165 |
+
headers={"Content-Disposition": "attachment; filename=result_cards.json"},
|
166 |
)
|
167 |
|
168 |
|
169 |
+
@router.get("/health", summary="Health check")
|
170 |
+
async def health_check():
|
171 |
+
return {"status": "healthy"}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
172 |
|
173 |
|
174 |
+
@router.get("/version", summary="Service version")
|
175 |
+
async def version_check():
|
176 |
+
return {"version": "1.0.0"}
|