def addToStream(x):
try:
detectedProbabilities = x[0].tolist()
detectedClasses = x[1].tolist()
bloburl = x[2]
weatherCondition = x[3]
windSpeed = x[4]
isDone = x[5]
result = []
streamResult = []
for idx, prediction in enumerate(detectedClasses):
result.append(
[Labels[prediction], detectedProbabilities[idx] * 100])
for idx, label in enumerate(Labels):
if not idx in detectedClasses:
result.append([label, 0])
df = pd.DataFrame(result)
result = df.groupby(0)[1].mean()
for (columnName, columnData) in result.iteritems():
streamResult.append([columnName, columnData])
streamResult.append(['fileName', bloburl])
streamResult.append(['isDone', isDone])
streamResult.append(['weather', weatherCondition])
streamResult.append(['windSpeed', windSpeed])
redisgears.executeCommand('xadd', 'predictions', '*',
*sum(streamResult, []))
except:
xlog('addToStream: error:', sys.exc_info())
def getSecret(secretName):
try:
with open('/run/secrets/' + secretName) as f:
secret = f.readline()
redisgears.executeCommand('xadd', 'secret', '*', 'text', secret)
return secret
except:
xlog('getSecret: error:', sys.exc_info())
def passAll(x):
try:
redisgears.executeCommand(
'xadd', 'all', 'MAXLEN', '~', str(MAX_IMAGES), '*', 'image',
'data:image/jpeg;base64,' +
base64.b64encode(x['img']).decode('utf8'))
except:
xlog('passAll: error:', sys.exc_info()[0])
def addToStream(x):
# save animal name into a new stream
try:
redisgears.executeCommand(
'xadd', 'cats', 'MAXLEN', '~', str(MAX_IMAGES), '*', 'image',
'data:image/jpeg;base64,' + base64.b64encode(x[1]).decode('utf8'))
except:
xlog('addToStream: error:', sys.exc_info()[0])
def predictImage(x):
try:
detectedProbability = np.array([])
detectedClasses = np.array([])
bloburl = ''
if x['value']['image']:
image_data = io.BytesIO(x['value']['image'])
image = Image.open(image_data)
numpy_img = np.array(image)
resize_img = cv2.resize(numpy_img, (320, 320),
interpolation=cv2.INTER_LINEAR)
inputs = np.array(resize_img,
dtype=np.float32)[np.newaxis, :, :, :]
img_ba = bytearray(inputs.tobytes())
v1 = redisAI.createTensorFromBlob('FLOAT', [1, 320, 320, 3],
img_ba)
graphRunner = redisAI.createModelRunner('customvisionmodel')
redisAI.modelRunnerAddInput(graphRunner, 'image_tensor', v1)
redisAI.modelRunnerAddOutput(graphRunner, 'detected_boxes')
redisAI.modelRunnerAddOutput(graphRunner, 'detected_scores')
redisAI.modelRunnerAddOutput(graphRunner, 'detected_classes')
res = redisAI.modelRunnerRun(graphRunner)
res1 = redisAI.tensorToFlatList(res[0])
res2 = redisAI.tensorToFlatList(res[1])
res3 = redisAI.tensorToFlatList(res[2])
redisgears.executeCommand('xadd', 'result1', '*', 'text', res1)
redisgears.executeCommand('xadd', 'result2', '*', 'text', res2)
redisgears.executeCommand('xadd', 'result3', '*', 'text', res3)
deleteLowProbResult = []
for idx, prediction in enumerate(res2):
if (prediction < 0.5):
deleteLowProbResult.append(idx)
array_2d_rowcount = int(len(res1) / 4)
arr_2d = np.reshape(res1, (array_2d_rowcount, 4))
detectedBoxes = np.delete(arr_2d, deleteLowProbResult, axis=0)
detectedProbability = np.delete(res2, deleteLowProbResult)
detectedClasses = np.delete(res3, deleteLowProbResult)
imagename = x['value']['imagename']
connectionString = getSecret("azure_blob_secret")
blob = BlobClient.from_connection_string(
conn_str=connectionString,
container_name=ContainerName,
blob_name=imagename)
add_boxes_to_images(image, detectedBoxes, detectedClasses, blob)
bloburl = getBlobUrl(imagename, connectionString)
weatherCondition = x['value']['weather']
windSpeed = x['value']['windSpeed']
isDone = x['value']['isDone']
return detectedProbability, detectedClasses, bloburl, weatherCondition, windSpeed, isDone
except:
xlog('Predict_image: error:', sys.exc_info())
def add_boxes_to_images(img, predictions, classes, blob):
try:
for idx, pred in enumerate(predictions):
x = int(pred[0] * 600)
y = int(pred[1] * 600)
width = int(pred[2] * 600)
height = int(pred[3] * 600)
shape = [(x, y), (width, height)]
redisgears.executeCommand('xadd', 'env', '*', 'text',
os.environ['FontPath'])
font = ImageFont.truetype(r'/data/fonts/ariblk.ttf', 20)
text = Labels[classes[idx]]
ImageDraw.Draw(img).rectangle(shape, outline="red")
ImageDraw.Draw(img).text((x, y),
text,
fill="red",
align="left",
font=font)
saveImageToAzure(img, blob)
except:
xlog('add_boxes_to_images: error:', sys.exc_info()[0])
def passAll(x):
redisgears.executeCommand(
'xadd', 'all', 'MAXLEN', '~', '1000', '*', 'image',
'data:image/jpeg;base64,' + base64.b64encode(x['img']).decode('utf8'))
def addToStream(x):
# save animal name into a new stream
redisgears.executeCommand(
'xadd', 'cats', 'MAXLEN', '~', '1000', '*', 'image',
'data:image/jpeg;base64,' + base64.b64encode(x[1]).decode('utf8'))
def xlog(*args):
redisgears.executeCommand('xadd', 'log', '*', 'text',
' '.join(map(str, args)))