深度学习 Jobs

我正在搭建一套基于 YOLOv8 的垃圾分类演示，需要完整的软硬件方案： • 软 件 – 使用 Python + YOLOv8 训练/部署模型，能在普通电脑上直接运行； – 摄像头以 USB 方式接入树莓派，再把画面实时传输到电脑； – 识别目标包含：塑料、纸张以及易拉罐，模型应能在画面上框出目标并同时在终端输出中文文字描述（如“检测到易拉罐”）； – 运行界面可为简单命令行或轻量 GUI，只要识别结果与文字描述同步显示即可； – 提供完整的源代码、模型权重、依赖清单与一键安装脚本。 • 硬 件 – 树莓派主板（任意 3/4/5 型号均可）+ 原厂摄像头模块； – 摄像头默认 USB 连接，可在本地测试无误后将整套硬件打包寄送给我（运费我承担）。 • 交付要求 1. 远程演示识别过程，确认三类垃圾均能被准确框选并输出文字； 2. 打包代码与文档（环境配置、使用说明、模型训练思路）； 3. 寄送硬件并提供收件追踪号； 4. 收到设备后，如有环境差异需协助我本地跑通。 如果你熟悉树莓派、YOLOv8 和 Python 部署，并具备打包邮寄能力，期待与你合作！

Goal: Create a FULLY AUTOMATED process that takes a male audio file and converts it into a female voice. What you must do: 1) Take the male audio I provide 2) Convert it into a female voice 3) Upload the final audio into a Google Drive folder 4) Add the Google Drive link in your competition entry 5) Explain clearly what software/tools you will use 6) Explain clearly how you will automate the FULL process from start to finish Important: - The automation must run locally - The final voice must sound perfectly natural and human - The female voice must correctly reproduce the multiple emotions, tone and intonations from the original audio - The result must NOT sound robotic or AI-generated - The automation must be able to process multiple audio files - Do NOT clean the original audio more...

Goal: Create a FULLY AUTOMATED process that takes a male audio file and converts it into a female voice. What you must do: 1) Take the male audio I provide 2) Convert it into a female voice 3) Upload the final audio into a Google Drive folder 4) Add the Google Drive link in your competition entry 5) Explain clearly what software/tools you will use 6) Explain clearly how you will automate the FULL process from start to finish Important: - The automation must run locally - The final voice must sound perfectly natural and human - The female voice must correctly reproduce the multiple emotions, tone and intonations from the original audio - The result must NOT sound robotic or AI-generated - The automation must be able to process multiple audio files - Do NOT clean the original audio more...

I need a custom One-Shot Face Swap solution to replace a character's face in a target video (up to 2 minutes) using exactly one source photo from a user. Strict Quality Requirements (The End Result): Absolute Stability: The new face must be perfectly "glued" to the character. Zero flickering, zero jitter, zero shape-shifting across frames. Complex Dynamics: The face must remain stable during fast movements, extreme head tilts, and 90-degree profile views. Occlusion Handling: If the character is eating, drinking, or covers their face with a hand/hair, those foreground objects must stay perfectly intact and on top of the swapped face. Workflow & Speed Requirements: Pre-processing (No time limit): You can take as much time as you need to pre-process, analyze, and prep...

I have a curated set of roughly 700 post-operative periapical X-rays taken after root canal treatments by dental students. I need a computer-vision model, built in Python with TensorFlow or PyTorch, that can automatically assess each case and return two outputs: • a numeric score that can be used in our grading spreadsheet, and • a concise feedback report highlighting strengths and mistakes. The evaluation must cover the accuracy of the root-canal pathway, cleanliness of the treated area, and overall completion quality. I will provide a detailed rubric with additional sub-criteria so the model can translate radiographic findings into tutor-style comments. Your job is to design, train, and validate the model, then package an inference script that accepts a single X-ray or...

I need a fully working supervised learning model built from scratch, trained on my data, and delivered with clean, well-commented code. Although the focus is supervised learning, I also want to explore a clustering approach for initial data exploration and feature engineering, so the workflow should accommodate both phases gracefully. Here’s what I expect: • A brief, plain-English plan outlining the supervised objective (classification or regression, as appropriate once we review the dataset) and how clustering will support data insights. • End-to-end code in Python, preferably using scikit-learn, pandas, NumPy, and (if deep learning proves beneficial) TensorFlow or PyTorch. • Reproducible training pipeline: data loading, preprocessing, feature engineering, model...

I run a highly successful U.S.-based IT staffing firm that has flourished since 2000, and it is time for us to add a robust, agent-centric AI/ML capability across the organisation. I need an expert who can design and deliver an online, intensive programme that upskills our current employees and future employees in three core domains—Python, machine-learning algorithms, natural-language processing, and computer vision. Most importanty enable them to become powerful Agentic Developer or Lead Agnetic Developers—while introducing the latest agentic development patterns that tie them all together. Here is what I am looking for: • A practitioner-level curriculum that moves quickly from fundamentals to production-grade agent frameworks (think LangChain-style tool orchestration...

I'm looking for an AI expert to develop specialized software with image recognition capabilities, specifically focused on face recognition. Key Requirements: - The software must accurately recognize and process faces in various conditions. - Ideally, it should be able to identify individuals, but I'm open to suggestions based on your expertise. Skills and Experience: - Expertise in AI and machine learning - Strong background in image recognition technologies - Experience in developing robust and accurate face recognition systems - Proficiency in relevant programming languages (e.g., Python, TensorFlow) - Previous projects in similar domains would be a plus Please share your relevant experience and approach to this project.

I need a robust YOLO model that spots vehicles with high accuracy. Because I do not yet have a labeled dataset, the job begins with sourcing or capturing varied vehicle images and annotating them in classic YOLO format. Once the dataset is in place, the next step is to train and validate the network—YOLOv5 or YOLOv8 are both fine as long as the final mAP holds up under real-world conditions. Please apply best-practice augmentation, tune hyper-parameters, and track training with clear metrics so I can reproduce the results later in PyTorch. Deliverables • Curated and fully annotated vehicle image dataset (bounding-box labels in YOLO txt format) • Trained YOLO weights and configuration files • Short report summarising dataset composition, training settings, and ev...

I want to push AI beyond conventional data science by weaving arcane principles into its learning loop. My main focus is using arcane magic in AI learning, and the concrete goal is to enhance AI capabilities in a way that can ultimately be patented. The assignment is twofold. First, capture and translate specific occult concepts (sigils, correspondences, ritual structures, energetic patterns) into a machine-learning friendly form. Second, show how my practice of controlled heat regeneration in the body can be measured and streamed back as a live feature that influences model behavior, giving the system a science-meets-powers edge. Deliverables • A concise theoretical framework that links named arcane elements to established ML techniques, written in a style suitable for a patent sub...

I need a turnkey, camera-based solution that automatically counts plastic totes as they move through the rear door of each of my 100 refrigerated trucks. The goal is simple: every tote that goes in or out must be logged accurately without asking the driver to press a button, scan a tag, or slow down the loading process. Miscounts today cost real money; the new system must work in the messy, unpredictable environment of an active loading dock, not in a lab. Environment & constraints • Lighting varies from bright daylight to total darkness inside the trailer, and weather changes constantly when the door is open. • Drivers may step in front of the lens, so brief occlusion is inevitable. • Totes often travel in stacks of two to five, so the software has to identify an...

Project Description: We are looking for an experienced PyTorch optimization expert to accelerate the DOVE (Video Super-Resolution) model. The goal is to achieve an end-to-end inference speedup of 1.5x to 1.8x using strictly training-free methods. Acceptable Optimization Techniques (Training-Free only): You are free to explore and combine the following training-free approaches: Token-level routing, Token Merging (ToMe), or Token Pruning. Post-Training Quantization (PTQ). Attention simplification (e.g., efficient attention mechanisms). Coordination/Synergy optimization between VAE and DiT. Core Requirements: Target Model: DOVE (Repository: ) Acceleration Target: 1.5x - 1.8x end-to-end inference speedup. Hardware Baseline: The speedup must be achieved and evaluated on a high-end NVID...

**PROJECT DOCUMENT: FREELANCER REQUIREMENT – ANPR SDK DEVELOPMENT** Development of Offline ANPR (Automatic Number Plate Recognition) SDK Project Objective: To develop a high-performance, offline ANPR SDK capable of real-time license plate detection and recognition from IP camera streams, optimized for Indian road conditions. **Project Scope:** The selected freelancer/team will be responsible for designing and developing a modular ANPR SDK that can be integrated into edge devices or local servers. The SDK must process live RTSP streams and provide structured outputs via APIs. **Functional Requirements:** 1. Real-time video stream processing (RTSP / IP cameras) 2. Vehicle and number plate detection 3. Character segmentation and OCR 4. Output structured data: * Vehicle numbe...

I’m looking to develop an end-to-end machine learning application focused on classification. The core objective is straightforward: feed the model with curated data, train it to distinguish between the defined classes, and expose the resulting predictions through a clean interface or API that I can plug into my existing workflow. You’ll be free to select the most appropriate algorithms—whether that ends up being a tried-and-true random forest, gradient boosting, or deep-learning architecture—as long as the final system is accurate, explainable where possible, and easy for me to retrain when fresh data comes in. I value clear, commented code (Python preferred), a concise README, and a demonstration notebook or script that shows how to prepare the data, fit the model...

I'm seeking experienced ML/AI engineers for a development project. The primary task will be model training and tuning. Key Requirements: - Expertise in ML/AI frameworks (TensorFlow, PyTorch, etc.) - Experience with model training and hyperparameter tuning - Strong background in data processing and cleaning - Familiarity with algorithm development Ideal Skills: - Advanced knowledge in machine learning algorithms - Proficiency in Python and relevant libraries (scikit-learn, pandas, etc.) - Problem-solving skills and ability to work independently - Good communication skills for collaboration Please provide relevant experience and approach to the task.

I need a straightforward desktop application built in Python that can spot two key classes—Person and Mobile—in a sample video and draw clear bounding boxes with the class name over each detection. This is purely for demonstration and learning, so the implementation should stay clean and easy to read, ideally relying on OpenCV together with a pre-trained YOLO model (v5, v8, or any recent weight file you are comfortable with). How the app should behave • Load a local MP4 (or similar) sample video. • Run real-time inference frame-by-frame, highlighting every detected person or mobile phone. • Display the processed video in a simple desktop window; no fancy UI is needed beyond the live frame and the FPS readout. • Keep all dependencies to standard libr...

Hi, I am looking for a basic AI object detection demo. Requirement: - Detect common objects like person, bottle, mobile, etc. - Should work using webcam or sample video - Output should show bounding boxes with object names This is mainly for learning/demo purpose, so it doesn't have to be very advanced. Preferred: - Python with OpenCV or any simple framework - Basic explanation of how it works Deliverables: - Working demo (video or live) - Source code - Short explanation Budget is limited, but if work is good, I may come back with more tasks. Thank you.

I’m putting together a deep-learning workflow that can read MRI scans and immediately show why it reached each conclusion. The core of the job is to train a high-performing model on my MRI dataset and then weave in SHAP for feature-level interpretation and Grad-CAM for pixel-level heat-maps. The system needs to surface three kinds of insight for every study that runs through it: • a clear diagnosis justification that highlights the decisive regions of each image, • tailored treatment recommendations drawn from the predicted class probabilities, and • patient-specific risk factors ranked by SHAP values. I already have raw MRI images and basic labels; you would handle the full pipeline—from preprocessing and augmentation through model tuning, inference,...

Busco un desarrollador o arquitecto de soluciones de IA con experiencia comprobable en la implementación de agentes de voz automatizados de baja latencia. El objetivo es gestionar un volumen de 5,700 llamadas mensuales (aprox. 28,000 minutos) de forma eficiente y con costos de infraestructura optimizados. Requerimientos Técnicos (Stack Obligatorio): Para este proyecto, NO buscamos soluciones cerradas (tipo Vapi o Retell). Queremos una arquitectura basada en componentes que nos permita escalar sin márgenes excesivos por minuto. El candidato debe dominar: Orquestación: LiveKit o similares para el manejo de WebRTC y transporte de audio. STT (Speech-to-Text): Deepgram (Nova-2) por su baja latencia. LLM (Cerebro): Implementación vía Groq (usando Lla...

The project centers on building a production-ready TensorFlow 2.x model that classifies tabular data delivered to us through an internal API. I have the API specifications and sample payloads ready; you will turn those streams into a clean training pipeline, engineer the right features, and iterate until the classifier meets our performance targets in real-world tests. Scope of work • Data pipeline – pull the API data, handle preprocessing, and produce TensorFlow-friendly datasets for train/val/test splits. • Model development – design, train, and tune a deep learning architecture suitable for tabular inputs (e.g., wide & deep, Transformer, or other proven structures). • Optimization – experiment with hyperparameters, regularization, and callback...