MNIST TUTORIAL#

Connect to EvaDB#

%pip install --quiet evadb
import evadb
cursor = evadb.connect().cursor()

[notice] A new release of pip is available: 23.0.1 -> 23.1.2
[notice] To update, run: pip install --upgrade pip

Note: you may need to restart the kernel to use updated packages.

Downloading https://github.com/ultralytics/assets/releases/download/v0.0.0/yolov8m.pt to yolov8m.pt...

  0%|          | 0.00/49.7M [00:00<?, ?B/s]

 10%|█         | 5.16M/49.7M [00:00<00:00, 54.1MB/s]

 51%|█████     | 25.4M/49.7M [00:00<00:00, 147MB/s]

 92%|█████████▏| 45.5M/49.7M [00:00<00:00, 176MB/s]

100%|██████████| 49.7M/49.7M [00:00<00:00, 163MB/s]

Download the video and load it into EvaDB#

# Getting MNIST as a video
!wget -nc https://www.dropbox.com/s/yxljxz6zxoqu54v/mnist.mp4

# Load the video into EvaDB
cursor.query("DROP TABLE IF EXISTS MNISTVid").df()
cursor.load("mnist.mp4", "MNISTVid", format="video").df()

File ‘mnist.mp4’ already there; not retrieving.

	0
0	Number of loaded VIDEO: 1

Run the Image Classification Function over the video#

# Connecting to the table with the loaded video
query = cursor.table("MNISTVid")

# Here, id refers to the frame id
# Each frame in the loaded MNIST video contains a digit
query = query.filter("id = 30 OR id = 50 OR id = 70 OR id = 0 OR id = 140")

# We are retrieving the frame "data" and 
# the output of the Image Classification function on the data 
# ("MnistImageClassifier(data).label")
query = query.select("data, MnistImageClassifier(data).label")

response = query.df()

2023-06-08 01:26:32,405	INFO worker.py:1625 -- Started a local Ray instance.

Visualize output of query on the video#

# !pip install matplotlib
import matplotlib.pyplot as plt
import numpy as np

# create figure (fig), and array of axes (ax)
fig, ax = plt.subplots(nrows=1, ncols=5, figsize=[6,8])

for axi in ax.flat:
    idx = np.random.randint(len(response))
    img = response['mnistvid.data'].iloc[idx]
    label = response['mnistimageclassifier.label'].iloc[idx]
    axi.imshow(img)
    
    axi.set_title(f'label: {label}')

plt.show()

../../_images/acebad3f1a6627edba0e8c3a9dff007a5fd3cf20992e7e95670726d402989427.png