PAICITY / SENSOR-TO-MODEL EXPLORATION

See what a sensor change does to your model.

A city workbench for exploring camera placement, sensor conditions and the evidence behind a model comparison. Keep the scene question concrete: what changed, what did the model receive and what happened to its output?

Actual PAICITY workbench · Ra’anana context. Assumed building heights are explicitly labelled and are display only.

CONTROL THE SENSOR CHANGE

Build a comparison you can inspect.

Move from a city scene to a paired experiment without losing sight of the input delivered to the model.

01 / Choose the scene

Explore a supported area and frame the operating question. The pilot areas include Rothschild Boulevard, Ramat Aviv, Haifa’s German Colony and Ra’anana’s Infinity Park.

02 / Configure the sensor

Change camera mounting and supported sensor parameters. Introduce controlled input degradations and compare the reference and changed views.

03 / Connect a model

Use the available local detector or configure a compatible external model endpoint through an adapter. VLM and VLA evaluation requires a connected model and a matching input/output contract.

04 / Inspect the evidence

Review paired inputs and recorded outputs. Export the available experiment evidence and replay metadata, then revise the setup and compare again.

MODEL RELIABILITY LAB

Keep the input beside the result.

A model score is easier to interpret when you can inspect the sensor input that produced it. Compare reference and changed frames alongside the experiment configuration.

Actual PAICITY evidence view · a recorded local-detector comparison. This capture displays historical evidence, not a newly executed VLM or VLA inference.

SENSOR CONFIGURATION

Ask a specific question.

How does a new mount position change visibility? What happens when the input is degraded? Keep each comparison tied to a controlled sensor change.

MODEL CONNECTION

Bring a compatible endpoint.

Adapters provide the connection surface for external models. A connected endpoint and an actual run are required before claiming model-specific results.

READ THE SCENE WITH ITS ASSUMPTIONS

Context and evaluation geometry stay explicit.

Urban context makes a scene easier to understand. Approximate context buildings and assumed heights are labelled as illustrative and excluded from sensor and physics evaluation.

Use PAICITY to investigate sensor-to-model questions. Use AIME when the question concerns mission execution, motion constraints or generated guardrails.

BRING YOUR ENGINEERING QUESTION

Start with a mission. Leave with evidence.

Explore the workspaces, then discuss a focused pilot around your mission, operating constraints or sensor configuration.