Performance & cost
This session
- Delivery time · median
- —
- Live model compute
- —
- Live input tokens
- 0
- Est. cost · USD
- $0.00
Free examples. Switch on to measure live inference with your key and quota.
Measurement details · 0 successful requests
- Successful requests
- 0
Run an example for free to see its measurements.
Model compute is the average per measured request. Delivery time is the median browser time for successful requests, including queue, network, and retries. Cached delivery speed is not inference speed. Model compute can exceed browser time when one request processes many inputs.
Tokens come from response headers. Inference costs use $0.04 per million input tokens, before plan credits. The estimate includes live and recorded tokens at the same rate; cached examples are not charged again. Missing usage is unavailable or a partial total (≥). Demo resets keep these totals; reload to start a new session.
Pick a photo and classify it.
Build with the SDKTypeScript · Python · dm1
One request from this demo. Copy it into your app.
Uses your MS_API_KEY and quota. Keep the key on your server.
npm install @cloudraker/milliseconds
export MS_API_KEY="your-api-key"import { readFile } from "node:fs/promises";
import { DecisionMachine } from "@cloudraker/milliseconds";
// Run on your server; reads MS_API_KEY from the environment.
// The image carries the input; the leading text is optional context.
const dm = new DecisionMachine();
const { result, usage } = await dm.classify(
await readFile("image.jpg"),
{
"crack": "A concrete surface with a visible crack: a narrow or wide irregular fracture line, possibly branching.",
"no_crack": "A concrete surface without a visible fracture line. Ordinary texture, pores, isolated pits, stains and lighting variation are not cracks."
},
{ detail: "low" }
).withUsage();
console.log(result);
console.log({ inputTokens: usage.inputTokens, modelMs: usage.inferenceMs });pip install cloudraker-milliseconds
export MS_API_KEY="your-api-key"from pathlib import Path
from milliseconds import DecisionMachine
# Reads MS_API_KEY from the environment.
# The image carries the input; the leading text is optional context.
dm = DecisionMachine()
result = dm.classify(
Path("image.jpg").read_bytes(),
{
"crack": "A concrete surface with a visible crack: a narrow or wide irregular fracture line, possibly branching.",
"no_crack": "A concrete surface without a visible fracture line. Ordinary texture, pores, isolated pits, stains and lighting variation are not cracks."
},
detail="low"
)
print(result)npm install -g @cloudraker/milliseconds
export MS_API_KEY="your-api-key"# Reads MS_API_KEY. The image and the other fields are read from disk.
cat > classify-labels.json <<'DM1_LABELS'
{
"crack": "A concrete surface with a visible crack: a narrow or wide irregular fracture line, possibly branching.",
"no_crack": "A concrete surface without a visible fracture line. Ordinary texture, pores, isolated pits, stains and lighting variation are not cracks."
}
DM1_LABELS
dm1 classify --image image.jpg --detail low --labels @classify-labels.json --json --usageUnder the hood How it works & limitations
API calls/classify
A first pass through inspection photos
A maintenance team can use a visual classifier to prioritize photos for inspection. The request sends one concrete crop to /classify with two described labels: visible crack or no visible crack.
Surface screening has a boundary
The dataset contains small surface crops. It cannot establish structural condition, crack depth, cause or severity. A negative result does not establish safety, and an inspector must make maintenance decisions. Pits, stains and shadows can also confuse a visual classifier.
Published inspection data
The eight unchanged sample JPEGs are from Concrete Crack Images for Classification, version 2 by Çağlar Fırat Özgenel, licensed CC BY 4.0. Positive source folders map to “crack”; Negative folders map to “no_crack”.
Dataset citation: Özgenel, Ç. F. (2019), Concrete Crack Images for Classification, Mendeley Data, V2, DOI: 10.17632/5y9wdsg2zt.2. Related study: Özgenel and Gönenç Sorguç (2018), Performance Comparison of Pretrained Convolutional Neural Networks on Crack Detection in Buildings, ISARC 2018.