azarnyx

from skimage import measure
labels = measure.label(mask)
props = measure.regionprops(labels, img_hed)[0]
angle = props.orientation

azarnyx

from skimage import morphology
mask = morphology.remove_small_objects(mask, 40000)
mask = morphology.remove_small_holes(mask, 40000)

azarnyx

img_rgb = skimage.io.imread('../pics/video_red/00535.jpeg')
mask = ((img_rgb[:, :, 0] > 130) & (img_rgb[:, :, 1] < 60) & (img_rgb[:, :, 2] < 60))

azarnyx

...
    img_rgb = skimage.io.imread(fnamei)

    # 1. RGB to HED
    img_hed = rgb2hed(img_rgb)
    img = img_hed[:, :, 1]

    # 2. Create mask using threshold
    mask = (img > 0.05)

azarnyx

...
for time in tqdm(ran):
    o0 = np.interp(time, Time, df['gyroRotationX(rad/s)'])
    o1 = np.interp(time, Time, df['gyroRotationY(rad/s)'])
    o2 = np.interp(time, Time, df['gyroRotationZ(rad/s)'])

    a0 = np.interp(time, Time, df['accelerometerAccelerationX(G)'])
    a1 = np.interp(time, Time, df['accelerometerAccelerationY(G)'])
    a2 = np.interp(time, Time, df['accelerometerAccelerationZ(G)'])

azarnyx

import pandas as pd
import pylab as plt
import numpy as np
from tqdm import tqdm

df = pd.read_csv("../data/measurements_new.csv")

# 0. Initialize variablles
e0 = np.array((1, 0, 0), np.longdouble)
e1 = np.array((0, 1, 0), np.longdouble)

azarnyx

import boto3
import json
client = boto3.client('sagemaker-runtime')
response = client.invoke_endpoint(
    EndpointName=predictor.endpoint_name,
    Body=json.dumps({"text": "preved medved"}),
    ContentType='application/json',
    Accept="application/json"
)
result = response['Body'].read().decode()

azarnyx

model.fit()
predictor = model.deploy(initial_instance_count=1, instance_type='ml.m5.4xlarge')

azarnyx

from sagemaker.predictor import RealTimePredictor, json_deserializer
import sagemaker
from sagemaker.pytorch import PyTorch, PyTorchModel
from sagemaker.sklearn import SKLearn
role = sagemaker.get_execution_role()

## Class to accept output in JSON format
class Predictor(RealTimePredictor):
    def __init__(self, endpoint_name, sagemaker_session):
        super().__init__(endpoint_name, sagemaker_session=sagemaker_session, serializer=None, 

azarnyx

# Specify that endpoint accept JSON
JSON_CONTENT_TYPE = 'application/json'

def predict_fn(input, model):
    proba = model.predict_proba(input)
    return json.dumps({
        "proba": str(list(proba[0]))
    })

def model_fn(model_dir):