roboflow-training-and-evaluation

name	roboflow-training-and-evaluation
description	Use when training Roboflow models, improving accuracy, or setting up a production feedback loop — covers architecture selection, model IDs, checkpoints, evaluation metrics, the iterative improvement playbook, and Active Learning through the Project Model Workflow block.

For agents — source-of-truth: This skill is authored in roboflow/computer-vision-skills and shipped with the Roboflow plugin. If your client has loaded the plugin (you'll see roboflow:<name> skills in your available skills list), use those local skills — they're read fresh from disk every session. The same content served as MCP resources at roboflow://skills/<name>/... is a fallback for clients without the plugin and may lag this repo. Don't call ReadMcpResourceTool for roboflow://skills/... URIs when a local roboflow:<name> skill is available.

Training & Evaluation on Roboflow

Training Flow

Upload/Annotate Images
  → Generate Dataset Version (preprocessing + augmentation + train/val/test split)
    → Pick Model Architecture + Size
      → Pick Checkpoint (COCO, Universe model, or previous version)
        → Train
          → Evaluate (auto-runs for paid users)

Version = frozen snapshot. Changes to the project after version creation do not affect it. Configure preprocessing (resize, contrast, etc.) and augmentation (flip, rotate, mosaic, etc.) during version generation.

Available Model Architectures

Object Detection

Architecture	Sizes	Default Resolution	Notes
RF-DETR	Pico, Nano, Small, Base, Medium, Large, XL, 2XL	384-880 (varies by size)	Best accuracy among named sizes; see RF-DETR NAS
Roboflow 3.0	Fast, Accurate, Medium, Large, XL	640x640	YOLOv8-based. Medium+ require paid plan
YOLO26	n/s/m/l/x	640x640	Also supports seg + pose
YOLOv12	n/s/m/l/x	640x640	OD only
YOLOv11	n/s/m/l/x	640x640	Also supports seg + pose
YOLOv8	n/s/m/l/x	640x640	Also supports seg + pose
YOLO-NAS	Small, Medium	640x640
YOLOLite CPU	n/s/m/l/x	640x640	Edge-optimized, beta
YOLOLite GPU	n/s/m/l/x	640x640	Edge-optimized, beta
Roboflow Instant	single	N/A (no resize)	Few-shot, free, OD only

Instance Segmentation

Architecture	Sizes	Default Resolution
RF-DETR Seg	Nano, Small, Medium, Large, XL, 2XL	312-768 (varies)
Roboflow 3.0 Seg	Fast, Accurate, Medium, Large, XL	640x640
YOLO-seg	v8/v11/v26 (n/s/m/l/x each)	640x640
SAM 3 (Segment Anything 3)	Large	1008x1008

Semantic Segmentation

Architecture	Sizes	Default Resolution
DeepLabV3+	Base	>=512x512

Classification

Architecture	Sizes	Default Resolution
ViT	Base	224x224
ResNet	18/34/50/101	224x224
DINOv3	Base, Small	224x224

Keypoint / Pose

Architecture	Sizes	Default Resolution
YOLO-pose	v8/v11/v26 (n/s/m/l/x each)	640x640

Multimodal / VLM

Architecture	Sizes	Default Resolution
Qwen3.5	0.8B, 2B	448x448
Qwen3 VL	2B	448x448
SmolVLM	256M, 2B	384x384
Florence 2	Base, Large	768x768
PaliGemma 2	3B	448x448
Qwen2.5 VL	7B	448x448

Model Selection Decision Tree

Follow this flowchart to pick the right model. Start at Step 1.

1. Task type? OD / Instance Seg / Keypoint → Step 2. Classification / Semantic Seg / VLM → use specialized block directly.
2. Target classes in COCO 80? Yes → Step 3. No → Step 6.
3. Real-time? No (images/recorded video) → Step 4. Yes (live video) → Step 5.
4. Non-real-time, COCO — OD / Inst Seg → prefer RF-DETR NAS; if unavailable, RF-DETR (detection) or RF-DETR Seg (instance segmentation) at Medium, Small for constrained HW, XL for accuracy-first. Keypoint → YOLO26 pose. Done.
5. Real-time, COCO — Same, and NAS is the strongest option here because it reports measured latency per target hardware. Without it, same families at Nano–Small. Done.
6. Non-COCO, which sub-task? OD → Step 7. Inst Seg → Step 8. Keypoint → Step 9.
7. OD, non-COCO — Check Rapid exclusions (see below). If excluded → Step 13. Otherwise → recommend Roboflow Rapid (default) or SAM3 zero-shot as secondary option → Step 10.
8. Inst Seg, non-COCO — SAM3 zero-shot (sam3/sam3_final, set class_names). Rapid does not support segmentation → Step 10.
9. Keypoint, non-COCO — Not real-time → YOLO26 pose Medium–XL. Real-time → YOLO26 pose Nano–Small. Done.
10. Real-time? No → try model (Step 11). Yes → warn about latency, try model (Step 12).
11. Non-real-time trial — User confirms works → Done. Poor results → Step 13.
12. Real-time trial — User confirms works → Done. Poor results → Step 13.
13. Universe Model Search — search community models on Roboflow Universe. Good match → Done. No match → Step 14.
14. Custom Training — OD / Inst Seg → prefer RF-DETR NAS. If NAS is unavailable, fine-tune RF-DETR (detection) or RF-DETR Seg (instance segmentation), sized by HW constraints. Done.

Model ID Reference

Read model-ids.md before calling a training tool. It contains the exact supported model_id values and the COCO class reference. Do not guess model IDs.

Model Selection Quick Guide

When you are training a model on the user's data for detection or instance segmentation, start with NAS. Neural Architecture Search searches the RF-DETR space against that data and reports a speed/accuracy frontier, so it is the option most likely to land on the best model for it. Reach for a single named architecture when NAS is unavailable (see prerequisites below), when the user asks for a specific one, or when a quick throwaway baseline is all that's wanted.

This summarises the training branches of the decision tree above; it does not override it. The tree may route a non-COCO request to Roboflow Rapid or SAM3 zero-shot first, neither of which trains a model — NAS only applies once custom training is the chosen path.

Goal	Recommended
Object detection — no strong prior	RF-DETR NAS (rfdetr-nas-parent)
Instance segmentation — no strong prior	RF-DETR NAS Seg (rfdetr-nas-seg-parent)
Best accuracy, object detection, NAS unavailable	RF-DETR (Large or XL)
Fast inference, object detection, NAS unavailable	RF-DETR Nano or YOLOv11n
Best accuracy, instance segmentation, NAS unavailable	RF-DETR Seg
Quick proof-of-concept (<1000 images)	Roboflow Instant
Classification	ViT or DINOv3
Multimodal / text prompts	Qwen3.5 or SmolVLM

NAS parents exist only for object detection and instance segmentation. Keypoint, classification, semantic segmentation, and VLM tasks have no NAS option — use the named models above.

Comparing architectures (sweeps)

If you are comparing architectures rather than picking one, include a NAS parent as one of the candidates whenever the prerequisites are met — e.g. rfdetr-medium vs yolo26m vs rfdetr-nas-parent. Launch one trainings_create per candidate and keep each trainingId.

Read the NAS arm's output as described under Reading a run in the RF-DETR NAS section below. One trap is specific to comparison: pick the representative to match the question, or you will understate NAS. The platform picks a winner per (metric, hardware) bucket by balancing accuracy against measured latency, and models_list exposes only the flattened union of those winners as a recommended boolean. So a flagged child won some bucket, which is not the same as being the most accurate: in one 76-model run the two flagged children scored 72.86 and 71.91 mAP50-95 while the best child scored 78.00, a 5–6 point gap. For a pure-accuracy comparison take the highest metrics.map5095 child. For a deployment decision, use the authoritative recommendedByHardware entry if the run exposes one; otherwise report the candidates' accuracy and latency for the target (reading latency per the shapes under Reading a run below), say that no per-hardware recommendation is exposed, and let the user pick the tradeoff. The run's own nasFamily: "baseline" children are a useful check on whether the search actually beat stock RF-DETR.

The NAS arm also takes longer than a single fine-tune, so report the named-model arms as they finish rather than blocking on NAS.

RF-DETR NAS (Neural Architecture Search)

Instead of picking a single RF-DETR size manually, NAS trains one parent model and mines many architectures out of it, reporting the speed/accuracy frontier so you can pick the one that fits your hardware budget.

• What: A NAS run trains a single parent model, then searches the RF-DETR architecture space within that trained parent to identify frontier candidates, reporting each one's mAP and measured latency on target hardware (e.g., Jetson, T4 GPU). The output is a set of models on a Pareto frontier, plus a winner auto-selected per (metric, hardware) bucket using Roboflow's current ranking heuristic to balance validation accuracy against measured latency. Those per-bucket winners are not exposed individually: models_list flattens them into one recommended boolean per child (see Picking for a specific hardware target).
• Tasks: Object Detection (rfdetr-nas) and Instance Segmentation (rfdetr-nas-seg).
• When to use: When you want the best speed/accuracy tradeoff for a specific deployment target and don't want to A/B-test sizes manually. Especially valuable for edge hardware where latency budgets are tight.
• Phases:
  1. Parent training — trains the one parent model the search draws from. This is the bulk of the wall-clock.
  2. Mining — searches architectures inside the trained parent and evaluates candidates to build the Pareto frontier (latency vs mAP). Candidates are derived from the parent rather than each being trained from scratch, so they appear in a burst near the end of the run. Each becomes a regular model you can deploy.
• Model IDs: rfdetr-nas-parent (Standard — use this by default), rfdetr-nas-pecoret-parent (Fast), rfdetr-nas-base-parent (Plus) for object detection; rfdetr-nas-seg-parent for instance segmentation.
• Prerequisites — check both before offering NAS:
  1. ≥15 validation images. versions_get returns splits.valid; below 15 the train call fails with insufficient_validation_images_for_nas, and waiting will not help — the user must generate a version with a larger validation split.
  2. Plan entitlement. NAS is included on Core and Growth plans; any other plan needs it granted on the workspace. Basic/starter/sandbox/research/trial need to upgrade; enterprise/legacy need to contact sales. Entitlement is not readable from the MCP, so this cannot be checked up front — a non-entitled workspace finds out when trainings_create rejects the run with code nas_not_available_for_plan. Treat that as a plan limit, not a transient error: do not retry it. Fall back to the named model for the task — rfdetr-medium (detection) or rfdetr-seg-medium (segmentation) — and say NAS is unavailable on the current plan and may need an upgrade or workspace enablement, using the plan on the error to tell which. Do not fall back to a hyperparameter sweep.
• Start a run: trainings_create(project_id, version_number, model_type="rfdetr-nas-parent"). NAS launches through the normal training tool; there is no separate engine parameter. (The UI equivalent is the Train page with ?engine=nas.) Results land at /{workspace}/{project}/nas-runs/{versionId}.
• Nothing to hand-tune: a NAS parent's whole hyperparameter surface is epochs (default 200, range 100–300). There are no learning-rate or loss-weight knobs, because the architecture search is the sweep.
• Reading a run: a run returns a frontier of models, not one — a 289-image dataset produced 76. Get modelGroup from trainings_list, which returns it without child metrics, then page the children with models_list(group=<modelGroup>, version_number=…, limit=…, offset=…). Don't use trainings_get for this: it inlines every child, which is the payload the paging exists to avoid. Each child carries a sparse metrics object. latency and paretoOptimalFor come from NAS mining, so they are on NAS children only - an ordinary training carries accuracy alone (e.g. map50, precision, recall, f1), and metrics can be null. Two shapes are in the data: newer runs report latency as a map keyed by hardware (e.g. {"AI1": 6.69, "T4": 1.98}) with paretoOptimalFor entries like "T4:map_50_95", older ones a scalar latency with a sibling metrics.hardware (e.g. "gpu") and bare entries like "map_50". Branch on the type rather than indexing. Children with nasFamily: "baseline" are stock RF-DETR models trained on the same data; they are never recommended and are a free within-run reference.
• Picking for a specific hardware target needs the authoritative map. The platform scores a separate winner per hardware, but models_list flattens that into one recommended boolean, true if the child won any bucket — so it cannot tell you which hardware, and paretoOptimalFor is unrelated frontier metadata, not the recommendation. The exact mapping is recommendedByHardware from trainings_get, but it is only built on the legacy version-based summary — a modern MMPV training returns no such field, which is the common case. So for most runs there is no exact per-hardware lookup at all. Do not infer one: report the candidates and ask which accuracy/latency tradeoff, or which latency budget, matters.
• Deploy: Each NAS-produced model deploys like any other — pick one and use it as a normal Roboflow model. Call it the hardware-recommended child only when an authoritative recommendedByHardware entry actually says so; otherwise it is the child the user chose from the frontier. Inference type is rfdetr-nas / rfdetr-nas-seg, but it's served through the standard inference paths.
• References: RF-DETR paper (arxiv), ICLR 2026, What is NAS? (blog).

Roboflow Instant / Rapid

Roboflow Instant

• What: Few-shot model, trains in minutes, free
• Task: Object Detection only
• When to use: PoC, <1000 images, quick iteration
• Auto-trains when you approve a batch and no Instant model exists yet
• No preprocessing/augmentation -- uses images as-is
• Deploy: Available in Workflows like any trained model
• Manual trigger: Project > Models > Train Model > Roboflow Instant Model

Roboflow Rapid

• What: Interactive annotation-and-training workflow — SAM3 pre-annotates a small image set, user reviews/corrects, a fast custom OD model trains automatically. Model keeps improving as it captures more production data.
• Task: Object Detection only, non-COCO classes
• When to use: Default path for non-COCO object detection when exclusions don't apply

Do NOT use Rapid when:

Exclusion	Why
OCR / text detection (characters, serial numbers, labels, receipts, license plates)	SAM3 cannot reliably segment individual characters
Blueprints, floor plans, schematics, technical drawings	Abstract symbols and line-based elements not handled by SAM3 text prompting
More than 5 target classes	SAM3 text prompting accuracy degrades significantly with many classes
Fine-grained visual distinctions (correct vs incorrect orientation, pass/fail, subtle defects)	SAM3 cannot differentiate nearly identical objects; fine-tuned model needed
High-precision measurement / metrology (distances, dimensions, tolerances)	SAM3 auto-labeling annotation precision insufficient for calibrated measurement

When Rapid is excluded → resume the decision tree at Step 13: Universe model search first, then custom training, which starts with RF-DETR NAS and falls back to named RF-DETR when its prerequisites are not met.

Checkpoint Training

Option	When to use
Public Checkpoint (COCO)	First model version, default recommended
Universe Checkpoint	Star a Universe project first, then it appears as checkpoint option. Good for domain-specific transfer learning
Previous Version	Already have a good model, want to improve with more data (all types except classification and SAM3)
Random Initialization	Advanced users only, usually worse results

Training Controls

• Cancel Training: Stops job, no weights saved. Refund if early in training.
• Early Stopping: Stops job, saves weights. Use when graphs show convergence with many epochs remaining. Charges for used credits.
• NAS Training: Shows paired charts (mining progress + Pareto curve, then per-model training curves). May auto-stop on convergence. See RF-DETR NAS section below.

Post-Training Metrics

Metrics vary by project type:

Project Type	Metrics Shown
Object Detection	mAP@50, Precision, Recall, F1
Classification	Accuracy
Instance Segmentation / Keypoint	mAP@50, Precision, Recall
Semantic Segmentation	mIoU
Multimodal	Perplexity

Model Evaluation (Paid Plans)

Auto-runs after training. Access: Models > click model version > View Evaluation.