Github

Overeasy

Dashboard

Support

Detections represent bounding box, segmentation, or classification results

Detection

👋 Introduction

🛠️ Installation

Install extra packages to speed up Overeasy

🤯 Extra Installations

Modal labs lets you run containerized Python code with GPUs.

📦 Modal setup

Workflows allow you to chain multiple visual agents.

🧩 Workflow

Execution Nodes represent individual steps in an image processing workflow.

🖼️ Execution Node

Execution Graphs manage and visualize the sequence of image processing tasks.

🔗 Execution Graph

Detections represent bounding box, segmentation, or classification results.

🔎 Detections

Discover Overeasy Agents: Specialized tools for every task.

🤖 Agents

The `BoundingBoxSelectAgent` detects an object in an image and selects its bounding box.

Bounding Box Select Agent

The `BinaryChoiceAgent` performs a binary classification.

Binary Choice Agent

The `ClassificationAgent` performs image classification.

Classification Agent

The `InstructorImageAgent` is designed to process images and extract structured data.

Instructor Image Agent

The `VisionPromptAgent` handles image-based queries.

Vision Prompt Agent

The `OCRAgent` performs Object Character Recognition (OCR) Tasks.

OCR Agent

The `DenseCaptioningAgent` generates detailed descriptions of images.

Dense Captioning Agent

The `TextPromptAgent` generates responses to text inputs that are augmented with a specific query.

Text Prompt Agent

The `InstructorTextAgent` is designed to process text inputs and extract structured data.

Instructor Text Agent

The `ClassMapAgent` is used to modify class names of detected objects.

Class Map Agent

The `NMSAgent` is used to perform Non-Maximum Suppression (NMS) on a set of detections, specifically for bounding boxes.

NMS Agent

The `ToClassificationAgent` is used to convert arbitrary data into a standardized classification format.

To Classification Agent

Join and Split Agents

⚙️ Models

DETIC

Grounding DINO

Open-World Language-guided Visual representation for object detection

OWLv2

Real-Time Open-Vocabulary Object Detections

YOLO-World

Contrastive Language-Image Pretrained Model

CLIP

Overview of OpenAI GPT models with a focus on image input capabilities