Beyond Generative AI Applications

SeekrFlow can also be used for general machine learning applications such as object detection, image classification, image segmentation, and text classification.

SeekrFlow provides an easy-to-use API to optimize your ML model to run on all supported platforms mentioned above. In this example, we will deploy a YOLO ("You Only Look Once") model for object detection in just 3 lines of code.

YOLO is a type of convolutional neural network for object detection tasks that is well-known for its speed and accuracy in detecting objects in real-time.

If you have a local YOLO model, feel free to use it, otherwise you can download an example trained on COCO from here:

wget https://github.com/ultralytics/assets/releases/download/v8.2.0/yolov8x.pt

import os
from seekrai import SeekrFlow

client = SeekrFlow(api_key=os.environ.get("SEEKR_API_KEY"))

model = client.models.upload("yolov8x.pt", model_type="object_detection")
client.models.promote(model.id)
client.models.predict(model.id, file="persons.jpg")

Here, we have fine-tuned YOLOv8 on the shipsnet from Kaggle. Let's try it on this image:

model = client.models.upload("best.pt", model_type="object_detection")
client.models.promote(model.id)
client.models.predict(model.id, file="ship.jpg")

# Get a list of all the image files in the training images directory
image_files = os.listdir(train_images)

# Choose 16 random image files from the list
random_images = random.sample(image_files, 16)

# Set up the plot
fig, axs = plt.subplots(4, 4, figsize=(16, 16))

# Loop over the random images and plot the object detections
for i, image_file in enumerate(random_images):
    row = i // 4
    col = i % 4

    # Load the image
    image_path = os.path.join(train_images, image_file)
    image = cv2.imread(image_path)

    # Load the labels for this image
    label_file = os.path.splitext(image_file)[0] + ".txt"
    label_path = os.path.join(train_labels, label_file)
    with open(label_path, "r") as f:
        labels = f.read().strip().split("\n")

    # Loop over the labels and plot the object detections
    for label in labels:
        if len(label.split()) != 5:
            continue
        class_id, x_center, y_center, width, height = map(float, label.split())
        x_min = int((x_center - width/2) * image.shape[1])
        y_min = int((y_center - height/2) * image.shape[0])
        x_max = int((x_center + width/2) * image.shape[1])
        y_max = int((y_center + height/2) * image.shape[0])
        cv2.rectangle(image, (x_min, y_min), (x_max, y_max), (0, 255, 0), 3)


    # Show the image with the object detections
    axs[row, col].imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    axs[row, col].axis('off')

plt.show()

Demote the model when finished using it:

client.models.demote(model.id)