Dakota Agros | Crop Analytics Portal

Imagery Selection

Orthomosaic TIFF

Preview Resolution

Map brightness 100%

Flight series & flowering notes

Draw the trial grid on a registered map, then use it as the reference. Each date is measured with its own imagery and point cloud.

Site-year (e.g. Fargo-2020)

Flight ID Acquisition / observation date Geometry version

Record percent of plants flowering

Count unique plants, not heads. Use the same footprint for numerator and denominator. A bloom/stand detection ratio is shown separately and can exceed 100%. Notes stay separate from model measurements.

Plot ID Measured footprint / rows Flowering definition Unique flowering plants Reference stand in this footprint Reference stand measurement date Evidence / observer / stand source

Notes are saved in this browser. Download the CSV for an independent backup.

Plot Delineator

Study / Experiment Name

Grid rows (north–south)

Grid columns (east–west)

Plot Order

Plots per replicate

Rows per Plot

Exclude Border Filler Rows/Cols

Top

Bottom

Left

Right

Row Spacing Buffer

15%

Column Spacing Buffer

15%

Analytics Setup

Detection method

Uses the same reviewed model pair as the Machine Learning portal on registered maps.

Open bloom confidence (0–1) Stand confidence (0–1)

PyTorch Model Weights

Detection Confidence Filter

10%

Canopy Plant Spacing (pixels)

Manual Digitization

Plot outlines Plot labels

Plant outlines Bloom outlines

Draft recovery and backups

Review edits are saved automatically in this browser. Use Save Shapes for permanent statistics and GIS exports.

Import draft backup

Target Crop

Attribute Type

Display Shape Type

Shapes: 0

Shapefile Export Name

De-noise deleted shapes

Load Saved Shapefile

Processing Orthomosaic...

Cropping spectral bands and running plant algorithms.

Setup Guide:

Double-click the four field corners. Plot 1 of replicate 1 starts at the southeast corner.

NDVI Scale

Min Max

Visual Overlay:

Plot ID	Row	Columns	NDVI	GNDVI	NDRE	OSAVI	ExG	VARI	TGI	Canopy Temp	CWSI	Flower Count	Stand Count	Bloom / stand ratio %
Generate a grid and click 'Run Crop Analytics' to view results.

Saved runs (Ctrl-click to select several; one revision per flight)

Planting date, optional Plot Trait

Load a saved study or generate flight statistics.

Green: measured value. Orange: partial observed count. Dashed lines are visual guides, not fitted growth curves. Open heads / stands is not biological flowering percentage. Count declines flag possible loss or missed detections, not confirmed mortality.

Date	Flight	Value	Count coverage	Comparison	Stand change flag

Local R analysis

Supply actual experimental blocks and treatment/genotype assignments. CSV columns: Plot_ID, Block, Treatment (or Genotype). Grid strips and chips are not biological replicates.

Experimental design CSV Analysis factor Model Single acquisition date Trait I verified the experimental design and group assignments.

Gaussian mixed-model templates require residual/variance checks, especially for counts and percentages. Inputs, exclusions, code, diagnostics and R session information are exported together.

Publication evidence and private backup

Builds a private archive with saved reviews, vectors, statistics, checksums and a plot-series figure. Coordinates and local source references remain in it. Nothing is published automatically.

Optional independent count reference CSV Reference source / independence Include the latest completed R analysis from this page (must match these selected observations).

Reference columns: Plot_ID, Flight_ID, Stand_Count and/or Flower_Count. Count errors do not measure object-level precision or recall.

· Restore an evidence backup to new versions

Demo flight

The anonymous demo contains only the selected plot crop, local crop coordinates and saved detections. It excludes geographic coordinates, trial identifiers, full source maps and weights. It is a separate editable demonstration copy.

Select Training Shapefiles Select All

Training Job Type

Base Weights to Fine-Tune

Learning Rate

Frozen Layers

Use Merge Strategy (Prevent Forgetting)

Epochs

Model Output File

Training Console Logs

Awaiting retraining task initiation...