RTMDet_PRODUCTION / tests /test_image.py
phuochungus's picture
change to yolo
aaa1b64
import os
import mmcv
import numpy as np
from app.routers.image import inference_image
import pytest
import pytest
from fastapi.testclient import TestClient
from fastapi.routing import APIRoute
from app.main import app
def endpoints():
endpoints = []
for route in app.routes:
if isinstance(route, APIRoute):
endpoints.append(route.path)
return endpoints
@pytest.fixture
def client():
client = TestClient(app, "http://0.0.0.0:3000")
yield client
@pytest.mark.skipif("/image" not in endpoints(), reason="Route not defined")
class TestImageRoute():
img = mmcv.imread('demo.jpg')
url = "http://0.0.0.0:3000/image"
def test_inferenceImage(self):
bboxes, labels = inference_image(mmcv.imread('demo.jpg'), 0.3, True)
assert len(bboxes.tolist()) > 0 and len(labels.tolist()) > 0 and len(bboxes.tolist()) == len(labels.tolist())
result = inference_image(self.img, 0.3, False)
assert type(result) is np.ndarray and result.shape == self.img.shape
def test_ImageAPI(self, client):
payload = {}
files=[
('file',('demo.jpg',open('demo.jpg','rb'),'image/jpeg'))
]
headers = {
'accept': 'application/json'
}
response = client.request("POST", "image", headers=headers, data=payload, files=files)
result = mmcv.imfrombytes(response.read())
assert response.status_code == 200 and result.shape == self.img.shape
def test_ImageAPI_one_channel_array(self, client):
np.zeros((1,640,640)).dump("one_channel.jpg")
payload = {}
files=[
('file', ("demo.jpg",open("one_channel.jpg", "rb"),'image/jpeg'))
]
headers = {
'accept': 'application/json'
}
response = client.request("POST", "image", headers=headers, data=payload, files=files)
assert response.status_code != 200
def test_ImageAPIWithThresHold(self, client):
payload = {}
files=[
('file',('demo.jpg',open('demo.jpg','rb'),'image/jpeg'))
]
headers = {
'accept': 'application/json'
}
response = client.request("POST", "image?threshold=1&raw=True", headers=headers, data=payload, files=files)
thresHold = 0.4
assert response.status_code == 200 # The result with threshold equal 0 is 0
# No detected object has 100% accuracy
assert len(response.json()["labels"]) == 0
payload = {}
files=[
('file',('demo.jpg',open('demo.jpg','rb'),'image/jpeg'))
]
headers = {
'accept': 'application/json'
}
response = client.request("POST", "image?threshold=" + str(thresHold) + "&raw=True", headers=headers, data=payload, files=files)
assert response.status_code == 200
for bbox in response.json()['bboxes']:
assert bbox[4] >= thresHold