Build a phenotyping workflow
Start with reproducible image analysis, add foundation-model segmentation, then benchmark on field data.
- PlantCV
- Meta SAM 3.1
- PhenoBench
A focused map of official courses, models, datasets, and tools for scientific machine learning, AI agents & LLMs, crop simulation, plant phenotyping, remote sensing, and 3D reconstruction.
Each path moves from a reliable starting point to a reproducible research workflow.
Start with reproducible image analysis, add foundation-model segmentation, then benchmark on field data.
Move from an approachable water-productivity model to Python workflows and full cropping-system simulation.
Read the physiology first, follow shared measurement protocols, then analyse gas-exchange data reproducibly in R.
Map the field, learn scientific machine learning, then reproduce a domain model with official notebooks.
Pick a dimension on the left—foundations, domain, or publishing—then read each topic from beginner to advanced.
Deep learning, graphs, classical ML, and the from-scratch explanations worth returning to.
Beginner
The official path from tensors and training loops to transfer learning, detection, distributed training, and compilation.
A careful introduction to tabular ML, evaluation, pipelines, model selection, and leakage-aware experimentation.
Build neural networks from scratch—backprop, makemore, and a GPT—through carefully narrated, code-along video lectures.
Visual, intuition-first explanations of neural networks, gradient descent, backpropagation, and transformers.
Complete written notes on regression, classification, neural networks, and reinforcement learning—useful when a video course moves too fast.
Intermediate
Hands-on graph neural networks for molecules, biological networks, meshes, point clouds, and heterogeneous graphs.
A durable computer-vision foundation covering recognition, optimization, CNNs, transformers, detection, and segmentation.
Advanced
A structured course on graph representation learning, GNNs, graph transformers, knowledge graphs, and applications.
In-depth technical essays on LLM agents, diffusion, RL, hallucination, and reasoning—widely used as reference explainers.
Model APIs, agent frameworks, tool protocols, and agricultural domain models.
Beginner
Official Claude developer docs: Messages API, tool use, extended thinking, structured outputs, prompt caching, and agent patterns.
Official OpenAI developer reference for the Responses API, function calling, structured outputs, embeddings, and retrieval.
A hands-on course on building agents with smolagents, LlamaIndex, and LangGraph, from fundamentals to a capstone project.
Intermediate
Build production agents on the same harness as Claude Code, with subagents, sessions, tool orchestration, and MCP support.
A lightweight framework for multi-agent workflows with handoffs, guardrails, sessions, and built-in tracing.
The open standard for connecting LLMs to tools and data sources, with a growing ecosystem of interoperable servers.
A minimal library for code-writing agents—define tools, pick any model, and run agentic loops in a few lines of Python.
The first seed-industry LLM, trained for variety selection, agronomic traits, cultivation, and promotion-region reasoning.
Advanced
A low-level orchestration framework for durable, stateful multi-agent systems with explicit graphs, memory, and human-in-the-loop.
An open Chinese agricultural multimodal model on MiniCPM-Llama3-V that diagnoses crop disease and answers farming questions.
A domain LLM ecosystem for agriculture built on a multi-agent data engine and the Agri-342K instruction dataset.
Photosynthesis, fluorescence, canopy measurement protocols, and how to analyse what comes off the instrument.
Beginner
An open plant-physiology textbook covering photosynthesis, water relations, growth, and stress—the background a modelling or imaging paper assumes you already have.
Community-maintained protocols for gas exchange, fluorescence, sensing, and environment measurement, written so a method can be repeated by someone else.
A handheld MultispeQ plus shared protocols and data, giving fluorescence and leaf measurements at field scale without a lab bench.
Intermediate
A practical Chinese walkthrough for fitting A/Ci and light-response curves and cleaning LI-COR output in R.
Manuals, application notes, and configuration guidance for the gas-exchange and fluorescence system most photosynthesis papers rely on.
How LAI is defined, measured directly and optically, and where each method's error comes from—useful before trusting an LAI product.
Advanced
Continuous field measurement of solar-induced fluorescence and reflectance—the ground reference behind canopy-scale SIF work.
Process-based crop simulation, from an approachable water model to full cropping systems and virtual plants.
Beginner
An approachable crop water-productivity model with official handbooks, reference manuals, and 43 video tutorials.
A multi-scale plant modelling initiative that shows how gene, cell, plant, and field models are coupled into one simulation.
Intermediate
A Python crop simulation environment with WOFOST, LINGRA, and LINTUL—well suited to optimization and data assimilation.
Modular C++ and R crop-growth simulation with practical guides for photosynthesis, environment, and model development.
Advanced
Model soil, water, nitrogen, crops, rotations, and management scenarios in a mature agricultural systems framework.
Learn genotype–soil–weather–management simulation for yield forecasting, cultivar calibration, and climate risk studies.
A declarative Julia framework for building, calibrating, evaluating, and visualizing crop and physiological models.
A C++ framework for 3D plant architecture with ray-traced radiation, energy balance, photosynthesis, and synthetic imagery.
A large-scale 3D radiative-transfer model that simulates reflectance, thermal imagery, LiDAR, and SIF over realistic canopies.
Image-to-trait workflows, segmentation models, field benchmarks, and the metadata standards behind them.
Beginner
The field's umbrella network: working groups, facilities, events, and a reliable map of who does what in phenotyping.
Intermediate
Reproducible RGB, NIR, thermal, fluorescence, hyperspectral, morphology, and geospatial plant-image workflows.
A focused workflow for PhenoCam time-series quality control, vegetation segmentation, indices, and phenology extraction.
Napari-based 2D and 3D cell segmentation tuned for densely packed plant tissues and microscopy workflows.
Python and C++ tutorials for point clouds, meshes, RGB-D data, registration, reconstruction, and 3D machine learning.
The common API for plant breeding and phenotyping data, so trials, germplasm, and observations move between systems without custom exports.
The minimum information checklist for a plant phenotyping experiment—apply it while collecting, not when a reviewer asks.
Advanced
A field benchmark for crop, weed, plant-instance, leaf-instance, and hierarchical panoptic segmentation from UAV imagery.
Multispectral 3D scans of four legume crops with organ-level leaf, petiole, and stem labels plus MIAPPE metadata.
Concept-prompted detection, segmentation, and tracking for images and video, with notebooks and fine-tuning code.
Visualize and quantify 4D live-imaged tissues, cell geometry, growth, fluorescence, and morphogenesis.
Spatio-temporal point clouds of maize and tomato with leaf instance labels—a rare benchmark for 4D plant growth analysis.
Earth observation platforms, geospatial foundation models, photogrammetry, and point-cloud workflows.
Beginner
Global crop extent and crop-type mapping with a MOOC, reference data, notebooks, and a processing hub.
A low-code workflow for mosaics, training samples, and RF/SVM/gradient-boosting land and crop classification.
The practical entry point to projections, vector and raster handling, and map layout before moving into scripted geospatial work.
Intermediate
Annual precomputed Earth representations for few-shot classification, clustering, land cover, and agricultural mapping.
Global field-boundary data, pretrained segmentation models, CLI, QGIS tooling, and browser inference.
Official JavaScript and Python learning paths for planetary-scale imagery, time series, classification, and export.
Turn drone imagery into orthophotos, point clouds, DSM/DTM, multispectral products, and textured 3D models.
Advanced
Fine-tune open geospatial foundation models for multi-temporal classification and segmentation, including crop mapping.
A lightweight self-supervised time-series transformer for Sentinel-1/2, weather, and terrain in low-label crop mapping.
A complete remote-sensing ML workflow using Sentinel-2, crop labels, cloud data engineering, and TensorFlow.
Scientific machine learning, physics-informed models, and domain foundation models with runnable code.
Beginner
A readable field map spanning graph learning, molecular simulation, causal ML, structural biology, and quantum science.
Intermediate
Notebook-first learning for molecular property prediction, drug discovery, quantum chemistry, and materials science.
Explore predicted protein structures and test biomolecular interactions with AlphaFold 3 through the official server.
A citation-grounded literature search and scientific question-answering workflow that can also use local models.
Advanced
A rigorous course on automatic differentiation, ODE/PDE solvers, PINNs, probabilistic programming, GPUs, and HPC.
Production-grade Physics AI tutorials for neural operators, PINNs, MeshGraphNet, weather, fluids, and molecular systems.
Current ESM code for protein language models, embeddings, structure prediction, and protein interaction research.
A foundation model for weather, air quality, waves, and tropical cyclones, with official ERA5 examples.
Manuscript language, controlled terminology, and figure galleries you can copy patterns from.
Beginner
Section-by-section sentence patterns for academic English—the fastest cure for a stalled introduction or discussion.
Reference-grade guidance on grammar, style, and citation formats, maintained by a university writing lab rather than a content farm.
Journal, thesis, CV, and poster LaTeX templates that compile in the browser—start from a working document instead of a blank preamble.
A decision tree from data type to appropriate chart, with the common mistakes each chart invites spelled out.
Intermediate
Reference ontologies for plant traits, anatomy, and growth stages—use these terms and your data stays searchable across species.
Hundreds of ggplot2 charts with reproducible code—the fastest way to go from a figure in your head to publishable output.
Editable interactive chart examples—useful when a figure needs to live on a web page or dashboard rather than in a PDF.