Implementing Object Detection with ImageAI in Just 10 Lines of Python

With only ten lines of Python, you can achieve state‑of‑the‑art object detection, a core task in computer vision that locates and identifies objects within images.

Object detection has become a fundamental component of artificial intelligence, evolving from early OpenCV algorithms to deep‑learning models such as R‑CNN, Faster‑R‑CNN, RetinaNet, SSD, and YOLO after the 2012 breakthrough in deep learning. The ImageAI library, created by Moses Olafenwa, abstracts these complex models so developers can apply them with minimal code.

To get started, install Python 3, then install the required packages via

pip install tensorflow numpy scipy opencv-python pillow matplotlib h5py keras

. Download the RetinaNet model file (e.g., resnet50_coco_best_v2.0.1.h5) and place it alongside your input image.

Below is the complete ten‑line script (save it as FirstDetection.py):

from imageai.Detection import ObjectDetection<br/>import os<br/><br/>execution_path = os.getcwd()<br/><br/>detector = ObjectDetection()<br/>detector.setModelTypeAsRetinaNet()<br/>detector.setModelPath(os.path.join(execution_path, "resnet50_coco_best_v2.0.1.h5"))<br/>detector.loadModel()<br/>detections = detector.detectObjectsFromImage(input_image=os.path.join(execution_path, "image.jpg"), output_image_path=os.path.join(execution_path, "imagenew.jpg"))<br/><br/>for eachObject in detections:<br/>    print(eachObject["name"] + " : " + eachObject["percentage_probability"])

The script imports ImageAI’s ObjectDetection class and Python’s os module, sets the working directory, configures the RetinaNet model, loads it, runs detection on image.jpg, saves the annotated result as imagenew.jpg, and prints each detected object’s name and confidence.

ImageAI also offers advanced features: you can adjust the minimum probability threshold, detect custom objects using the CustomObject class, choose detection speed levels (fast, faster, fastest), and specify input/output as file paths, NumPy arrays, or streams. By setting extract_detected_objects=True, the library extracts each detected object into separate image files.

Running the script produces before‑and‑after images that clearly show the detected objects with their names and probabilities, illustrating how a complex deep‑learning task can be simplified to a handful of lines of code.