此 MCP 可以做什么
Provides computer-vision calculations for image probing, bounding-box conversion, resizing, color contrast, detection metrics, and non-maximum suppression.
功能
工具
6
工具
bbox_convert
Bbox Convert
Convert bounding boxes between COCO, Pascal VOC and YOLO. FREE.
The three formats disagree on everything: COCO is [x, y, width, height],
VOC is [x1, y1, x2, y2], YOLO is [cx, cy, w, h] normalised to the image.
Getting this wrong produces boxes that look plausible and quietly ruin
every metric. Typical input {"boxes": [[10, 20, 100, 50]], "from_format":
"coco", "to_format": "yolo", "image_width": 640, "image_height": 480}
returns {"boxes": [[0.0938, 0.0938, 0.1562, 0.1042]], "converted": 1,
"rejected": []}.
Use whenever a dataset and a model disagree about format. Not for scoring
predictions (detection_metrics) and not for removing overlaps (nms). Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"} (for example {"error": "boxes must contain at least one box"}). Every call is read-only and idempotent, so after correcting the input it is always safe to retry.
只读
幂等
输入模式
{'type': 'object', 'required': ['boxes', 'from_format', 'to_format'], 'properties': {'clip': {'type': 'boolean', 'default': False, 'description': 'When true, clamp boxes to the image bounds instead of returning\nthem as they are. Off by default, because a box outside the image\nis usually a bug worth seeing.'}, 'boxes': {'type': 'array', 'items': {'type': 'array', 'items': {'type': 'number'}}, 'description': 'Boxes to convert, each a list of exactly four numbers in\nfrom_format, e.g. [[10, 20, 100, 50]].'}, 'to_format': {'enum': ['coco', 'voc', 'yolo'], 'type': 'string', 'description': 'The format to convert to; same three choices.'}, 'from_format': {'enum': ['coco', 'voc', 'yolo'], 'type': 'string', 'description': '"coco" for [x, y, w, h], "voc" for [x1, y1, x2, y2], or\n"yolo" for normalised [cx, cy, w, h].'}, 'image_width': {'type': 'integer', 'default': 0, 'minimum': 0, 'description': 'Pixel width, required whenever yolo is on either side.'}, 'image_height': {'type': 'integer', 'default': 0, 'minimum': 0, 'description': 'Pixel height, required whenever yolo is on either side.'}}, 'additionalProperties': False}
输出模式
{'type': 'object', 'additionalProperties': True}
colour_check
Colour Check
Check a colour pair against the WCAG contrast thresholds. FREE.
Uses the WCAG 2 relative-luminance formula, so the number matches what an
accessibility audit will report. Typical input {"foreground": "#767676",
"background": "#ffffff"} returns {"contrast_ratio": 4.54, "AA": true,
"AAA": false, "required": {"AA": 4.5, "AAA": 7.0}, "verdict": "Passes AA
for normal text, fails AAA."}.
Use when generating or auditing an interface. Not for converting colours
between spaces and not for palettes. Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"} (for example {"error": "foreground must be a hex colour like #767676"}). Every call is read-only and idempotent, so after correcting the input it is always safe to retry.
只读
幂等
输入模式
{'type': 'object', 'required': ['foreground', 'background'], 'properties': {'background': {'type': 'string', 'description': 'Background colour in the same notation.'}, 'foreground': {'type': 'string', 'description': 'Text colour as hex, e.g. "#767676" or "767676" or "#777".'}, 'large_text': {'type': 'boolean', 'default': False, 'description': 'True for text at least 18pt, or 14pt bold, which WCAG\nallows to pass at a lower ratio. Default false.'}}, 'additionalProperties': False}
输出模式
{'type': 'object', 'additionalProperties': True}
detection_metrics
Detection Metrics
Score detections against ground truth and show the working. PREMIUM (license).
Greedy matching at the IoU threshold, highest-confidence prediction first,
each ground-truth box matched at most once - the standard protocol.
Reports per-class precision, recall and F1, and average precision by the
all-points interpolation used by Pascal VOC 2010 onward. Typical input
{"predictions": [{"box": [0,0,10,10], "label": "cat", "score": 0.9}],
"ground_truth": [{"box": [1,1,11,11], "label": "cat"}]} returns
{"overall": {"tp": 1, "fp": 0, "fn": 0, "precision": 1.0, "recall": 1.0,
"f1": 1.0}, "per_class": {...}, "mAP": 1.0}.
Use to compare two models on the same held-out set. Not for cleaning up a
single model's overlapping output first - run nms before this. Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"} (for example {"error": "ground_truth must contain at least one box"}). Every call is read-only and idempotent, so after correcting the input it is always safe to retry.
只读
幂等
输入模式
{'type': 'object', 'required': ['predictions', 'ground_truth'], 'properties': {'box_format': {'enum': ['coco', 'voc', 'yolo'], 'type': 'string', 'default': 'voc', 'description': '"voc", "coco" or "yolo". Default "voc".'}, 'image_width': {'type': 'integer', 'default': 0, 'minimum': 0, 'description': 'Pixel width; required for yolo boxes.'}, 'predictions': {'type': 'array', 'items': {'type': 'object', 'additionalProperties': True}, 'description': 'Predicted boxes, each {"box": [...], "label": ...,\n"score": ...}. Score defaults to 1.0 when omitted.'}, 'ground_truth': {'type': 'array', 'items': {'type': 'object', 'additionalProperties': True}, 'description': 'True boxes, each {"box": [...], "label": ...}.'}, 'image_height': {'type': 'integer', 'default': 0, 'minimum': 0, 'description': 'Pixel height; required for yolo boxes.'}, 'iou_threshold': {'type': 'number', 'default': 0.5, 'maximum': 1, 'description': 'Overlap at which a prediction counts as a match.\nDefault 0.5, the usual reporting threshold.', 'exclusiveMinimum': 0}}, 'additionalProperties': False}
输出模式
{'type': 'object', 'additionalProperties': True}
image_probe
Image Probe
Read an image's format and pixel size from its header alone. FREE.
Dimensions live in the first few dozen bytes of PNG, JPEG, GIF, BMP and
WebP, so a base64 prefix is enough - you do not need to send the whole
file, and nothing is decoded. Typical input {"data_base64": "iVBORw0KG..."}
returns {"format": "png", "width": 1920, "height": 1080, "aspect_ratio":
1.7778, "aspect_label": "16:9", "megapixels": 2.07, "orientation":
"landscape", "bytes_inspected": 512}.
Use to find out what you are dealing with before planning a resize. Not for
pixel content - nothing here reads pixels - and not for EXIF. Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"} (for example {"error": "data_base64 must not be empty"}). Every call is read-only and idempotent, so after correcting the input it is always safe to retry.
只读
幂等
输入模式
{'type': 'object', 'required': ['data_base64'], 'properties': {'data_base64': {'type': 'string', 'description': 'The image file, base64-encoded. The first few hundred\nbytes are enough for every supported format; send a prefix rather\nthan a large file. Data-URL prefixes like "data:image/png;base64,"\nare accepted and stripped.'}}, 'additionalProperties': False}
输出模式
{'type': 'object', 'additionalProperties': True}
nms
Nms
Remove duplicate detections of the same object. PREMIUM (license).
Greedy non-maximum suppression: keep the highest-scoring box, drop
everything overlapping it above the threshold, repeat. Ties break on the
earlier index, so the result is deterministic rather than dependent on sort
stability. Typical input {"boxes": [[0,0,10,10],[1,1,11,11],[50,50,60,60]],
"scores": [0.9, 0.8, 0.7]} returns {"keep": [0, 2], "suppressed": [{"index":
1, "by": 0, "iou": 0.6807}], "kept": 2}.
Use after a detector that emits overlapping boxes. Not for scoring against
ground truth (detection_metrics) and not for format changes (bbox_convert). Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"} (for example {"error": "boxes and scores must be the same length"}). Every call is read-only and idempotent, so after correcting the input it is always safe to retry.
只读
幂等
输入模式
{'type': 'object', 'required': ['boxes', 'scores'], 'properties': {'boxes': {'type': 'array', 'items': {'type': 'array', 'items': {'type': 'number'}}, 'description': 'Candidate boxes in box_format, e.g. [[0, 0, 10, 10]].'}, 'scores': {'type': 'array', 'items': {'type': 'number'}, 'description': 'One confidence per box, same order and same length as boxes.'}, 'box_format': {'enum': ['coco', 'voc', 'yolo'], 'type': 'string', 'default': 'voc', 'description': '"voc", "coco" or "yolo". Default "voc".'}, 'image_width': {'type': 'integer', 'default': 0, 'minimum': 0, 'description': 'Pixel width; required for yolo boxes.'}, 'image_height': {'type': 'integer', 'default': 0, 'minimum': 0, 'description': 'Pixel height; required for yolo boxes.'}, 'iou_threshold': {'type': 'number', 'default': 0.5, 'maximum': 1, 'description': 'Overlap above which the lower-scoring box is dropped.\nDefault 0.5.', 'exclusiveMinimum': 0}}, 'additionalProperties': False}
输出模式
{'type': 'object', 'additionalProperties': True}
resize_plan
Resize Plan
Work out the exact scale, padding and crop for a model input size. FREE.
Returns the numbers you need to transform boxes alongside the image, which
is the step that usually gets skipped. Typical input {"width": 1920,
"height": 1080, "target": "yolo_640"} returns {"scale": 0.3333,
"resized": [640, 360], "pad": {"left": 0, "top": 140, "right": 0,
"bottom": 140}, "box_transform": "x_new = x * 0.3333 + 0; y_new = y *
0.3333 + 140"}.
Use before feeding an image to a fixed-input model. Not for finding out the
image's size in the first place - that is image_probe. Errors: on invalid, missing, or malformed input this tool never raises a protocol error — it returns {"error": "<what is wrong and how to fix it>"} (for example {"error": "unknown target <value>; use one of <value> or set"}). Every call is read-only and idempotent, so after correcting the input it is always safe to retry.
只读
幂等
输入模式
{'type': 'object', 'required': ['width', 'height'], 'properties': {'mode': {'enum': ['letterbox', 'cover', 'stretch'], 'type': 'string', 'default': 'letterbox', 'description': '"letterbox" scales to fit and pads the remainder, preserving\naspect; "cover" scales to fill and crops the overflow; "stretch"\ndistorts to fit exactly. Default "letterbox".'}, 'width': {'type': 'integer', 'description': 'Source image width in pixels.', 'exclusiveMinimum': 0}, 'height': {'type': 'integer', 'description': 'Source image height in pixels.', 'exclusiveMinimum': 0}, 'target': {'type': 'string', 'default': 'yolo_640', 'description': 'A named preset: "clip_224", "vit_384", "yolo_640", "sam_1024",\n"sd_512", "sd_768" or "detr_800". Ignored when target_size is set.'}, 'target_size': {'type': 'integer', 'default': 0, 'minimum': 0, 'description': 'A square side length in pixels, overriding target. Use\nthis for a size the presets do not cover.'}}, 'additionalProperties': False}
输出模式
{'type': 'object', 'additionalProperties': True}
历史
近期工具变更
已添加
detection_metrics
2026年9月17日 12:36
已添加
nms
2026年9月17日 12:36
已添加
colour_check
2026年9月17日 12:36
已添加
resize_plan
2026年9月17日 12:36
已添加
bbox_convert
2026年9月17日 12:36
已添加
image_probe
2026年9月17日 12:36
发现
类似的 MCP 服务器
kernelcad
com.kernelcad/kernelcad
Authors, evaluates, reviews, simulates, edits, and exports parametric CAD assemblies and geometry from editable TypeScript-based …
Affine.Earth Math Court
earth.affine/math-court
Performs exact integer and rational mathematical verification across algebra, geometry, physics, quantum-computing witnesses, rob…
RQM Studio Jobs
com.rqmtechnologies/studio
Provides bounded quantum-circuit parsing, validation, compilation, optimization, equivalence checks, repair, simulation, provider…
Bioregistry
io.github.pipeworx-io/bioregistry
Looks up biomedical identifier prefixes and metadata, including URI patterns, aliases, homepages, and cross-registry mappings.
RQM Jobs MCP
com.rqmtechnologies/jobs
Runs bounded robotics and control-system analyses such as trajectory validation, controller comparison, calibration assessment, c…
Mathjs
io.github.pipeworx-io/mathjs
Evaluates mathematical expressions and performs unit conversions through the Math.js API.
DOI Shape
io.github.sadri-dridi/doi-shape
Checks the structural format of a DOI.
Cords to cubic meters, number discarded
io.github.sadri-dridi/cord-m3
Provides cord-to-cubic-meter conversion along with small URL, parsing, time, weather, and validation utilities.