Steel Data Provider: From Crude-Output Tonnages to the Defect Mask · Head-to-head

Steel Plates Faults (UCI) vs JISF Japan steel statistics

Which steel data provider: from crude-output tonnages to the defect mask data fits your job: Steel Plates Faults, or JISF Japan steel statistics. API, files, or your warehouse. Daily, weekly, or hourly.

Steel Data Provider: From Crude-Output Tonnages to the Defect Mask Single unspecified steel producer · Static snapshot donated October 2010

Steel Plates Faults (UCI)

Steel Data Provider: From Crude-Output Tonnages to the Defect Mask Japan national totals · Monthly to July 2026 at review

JISF Japan steel statistics

Where the fields line up

No shared field names. These two answer different questions.

Field Steel Plates Faults JISF Japan steel statistics
X_Minimum Minimum X coordinate of the fault region on the plate. not in this set
X_Maximum Maximum X coordinate of the fault region on the plate. not in this set
Y_Minimum Minimum Y coordinate of the fault region on the plate. not in this set
Y_Maximum Maximum Y coordinate of the fault region on the plate. not in this set
Pixels_Areas Area of the fault region in pixels. not in this set
X_Perimeter Perimeter of the fault region along the X direction. not in this set
Y_Perimeter Perimeter of the fault region along the Y direction. not in this set
Sum_of_Luminosity Summed luminosity over the fault region pixels. not in this set
Minimum_of_Luminosity Minimum luminosity value within the fault region. not in this set
Maximum_of_Luminosity Maximum luminosity value within the fault region. not in this set
Length_of_Conveyer Length of the conveyor carrying the plate. not in this set
TypeOfSteel_A300 Binary flag indicating the plate is A300-type steel. not in this set

Coverage, side by side

Steel Plates Faults JISF Japan steel statistics
Geographic Single unspecified steel producer Japan national totals; trade detail for Korea, China, Taiwan, Thailand and the United States; companion series covering main producing countries
Temporal Static snapshot donated October 2010 Monthly to July 2026 at review; multi-decade annual and quarterly history; fiscal year April to March
Granularity One row per fault instance on an individual plate Monthly and annual aggregates by furnace type, product line or trade lane; quarterly rows in history files

What each contains

They tie on 1 attribute. Pick by fit, not by loyalty.

Steel Plates Faults JISF Japan steel statistics
Publisher M. Buscema, S. Terzi and W. Tastle, donated to the UC Irvine Machine Learning Repository Japan Iron and Steel Federation (JISF)
Subject lens Per-instance surface-defect metrology on steel plates with seven-class fault labels National steel statistics: production, products, order books, trade and inventories
Documented fields 34 fields - 27 predictors plus 7 one-hot labels - definitions verified during research 8 documented constructs, definitions verified during research
Geographic coverage Single unspecified steel producer Japan national totals; trade detail for Korea, China, Taiwan, Thailand and the United States; companion series covering main producing countries
Temporal coverage Static snapshot donated October 2010 Monthly to July 2026 at review; multi-decade annual and quarterly history; fiscal year April to March
Granularity One row per fault instance on an individual plate Monthly and annual aggregates by furnace type, product line or trade lane; quarterly rows in history files
Data shape 1,941 rows by 34 columns in one space-separated matrix, with a column-name sidecar file Roughly a dozen Excel workbooks per cycle, each with multiple product and period sheets
Secondary industries None None
Best for Training and benchmarking defect classifiers on identical labeled data Tracking Japanese steel supply, demand, trade and inventory cycles

Where they're equivalent

More than their subjects suggest. Same editorial standard: both field dictionaries were definition-verified during research rather than inferred, and both state their provenance explicitly - UCI names its three donors and the donation date, JISF names its next release date. Same machine-readable packaging philosophy: one ships a space-separated matrix with a column-name sidecar, the other ships Excel workbooks whose sheets parse directly. Same single-industry focus - both sit squarely in the steel slice with no secondary industries attached. And both are fixed-shaped in their own way: the UCI snapshot has not changed since October 2010, while each JISF workbook, once published, holds its month still - only the frontier advances.

The verdict

Verdict: sample both - they are not rivals, and your job decides which one opens first.

Take Steel Plates Faults (UCI) if the question is algorithmic. Training a defect classifier, comparing feature-selection schemes or class-imbalance fixes on identical data, building course material - the seven-label target over 27 numeric descriptors is a complete exercise kit, and the 2010 freeze date is what makes two teams' results comparable. It anchors the data-scientists reading of steel.

Take JISF Japan steel statistics if the question is commercial. Japanese crude steel momentum by furnace type, export pressure by destination, order-book direction by consuming sector, dealer inventory swings - anything an investors-and-quants reading of steel would act on lives in the monthly workbooks, refreshed on a schedule published in advance.

If the question spans both - whether mill-floor quality economics leave a trace in national output statistics - neither answers alone. That is a both-and request, and Datadory takes both-and requests.

Sample both, pick by fit. See Steel Plates Faults · See JISF Japan steel statistics

Or take both in one feed

Yes - as microscope and barometer rather than substitutes. A defensible workflow: let the JISF workbooks define the demand environment month by month (production, orders, trade, inventories), then reach for Steel Plates Faults when the analysis descends from market tone to unit-level quality - what a fault actually looks like in measurement space, and how separable the seven defect classes really are. Two seams decide whether the pairing holds. First, there is no join key: the UCI rows carry no plant, date or grade identifiers, so any link between the datasets is analytical analogy, never a literal merge. Second, respect the clocks: the plate samples describe 2010-era inspection practice while the JISF tables advance monthly - treat the pair as two zoom levels on one industry, not one longitudinal panel. Sampled on their own terms they stay honest; forced into a single table they mislead. And when the answer is simply yes to both, take them in one feed.

API, files, or your warehouse. Daily, weekly, or hourly.

Fair questions

Is Steel Plates Faults (UCI) better than JISF Japan steel statistics?

Better at different jobs. Steel Plates Faults wins for model-building: 1,941 labeled samples, 27 numeric features, seven defect classes and zero missing values make a complete classification exercise. JISF wins for market awareness: monthly crude steel output by furnace type, hot-rolled products, trade by destination and dealer inventories. One trains algorithms, the other briefs strategy.

Do the two steel datasets share any fields?

One construct: a numeric measure of steel. Steel Plates Faults measures a fault's pixel area; JISF measures crude steel output in thousand tonnes. Everything else diverges - 26 geometry-and-luminosity columns plus seven labels on one side, eight macro constructs spanning production, trade values and change rates on the other. There is no common key and no literal join between them.

Which dataset updates more often?

JISF, decisively. Its seven series publish monthly with the next release date announced in advance - July 2026 data landed mid-August, with September 24 named as the following slot. Steel Plates Faults is static: a snapshot donated in October 2010, unchanged since. Frozen is a virtue for benchmarks and a limitation for monitoring.

Can I combine Steel Plates Faults and JISF data in one analysis?

As two zoom levels, yes; as one merged table, no. The UCI rows carry no plant, date or grade identifiers, so nothing joins literally - the honest pattern uses JISF workbooks for the demand backdrop and Steel Plates Faults for unit-level fault geometry. Keep the clocks separate too: 2010-era inspection samples against live monthly statistics.

Can I get both steel datasets from Datadory?

Yes - sample both and pick by fit, or take both in one feed. Steel Plates Faults arrives as the full 1,941-by-34 labeled matrix; JISF arrives as its seven monthly series normalized to their documented fields, sample rows attached for inspection before anything ships. Delivered daily, weekly, or hourly - your call.