vizor

A fast detector kept honest by a slower, smarter one.

Install · Quick start · How it works · API · Write your own · Docs

A small detector runs on every frame but gets classes wrong. A vision-language model gets them right but is too slow for every frame, and often returns no boxes at all. Vizor runs both. Boxes and track ids come from the detector, class labels come from the VLM, and each answer is cached against the track id. A track the detector is confident about never reaches the VLM. One it is unsure about goes once, not once per frame.

Full documentation is at y-t-g.github.io/vizor. Everything below runs from this repo, and examples/ holds each snippet as a script you can run.

Install

The core needs numpy and OpenCV only. The model wrappers are optional extras, so install the one you actually use:

pip install vizor              # core
pip install "vizor[hf]"        # local Florence-2 or another transformers VLM
pip install "vizor[api]"       # hosted VLMs over OpenAI or Gemini

Model wrappers are imported on first use, so import vizor never pulls in torch if you are not using a torch model.

vizor ships no detector. examples/yolo.py is a working ultralytics adapter you copy, and it is not installed. See the header in that file.

Quick start

A detector on every frame, a hosted VLM on the boxes it is unsure of. YOLO comes from examples/yolo.py, not from vizor:

import vizor as vz
from yolo import YOLO

# YOLO detects and tracks every frame. The VLM only sees boxes YOLO scored at or
# below conf, cut out and sent up to eight per request. Each answer is cached
# against the track id, so an object costs one call however long it stays in view.
viz = vz.Vizor(
    YOLO("yolo11n.pt"),
    vz.VLM("gemini-3.1-flash-lite", api="gemini"),
    conf=0.5,
    mode="crop",
)

# run() yields the refined boxes for each frame and writes the annotated video
# as it goes. Drop save= and nothing is written.
for out in viz.run("traffic.mp4", save="out.mp4"):
    print(len(out), "boxes")

That writes an annotated out.mp4 and yields the refined boxes for every frame. The key is read from GEMINI_API_KEY, never passed in code.

Florence-2 instead, running locally and grounding the whole frame in one pass:

import vizor as vz
from yolo import YOLO

# the class list is Florence's prompt, so it looks for these three and nothing else
names = {0: "person", 2: "car", 7: "truck"}

# full mode grounds the whole frame in one pass, then matches Florence's boxes to
# the tracks by IoU. A match overwrites the box and confidence, not just the class.
viz = vz.Vizor(YOLO("yolo11n.pt"), vz.Florence(names=names), conf=0.5, mode="full")

# 0 is the first webcam. show=True opens a window, q or Esc closes it.
for out in viz.run(0, show=True):
    pass

Florence gives boxes as well as labels, so the refiner can correct the box too, not just the class.

To try it with no models at all, replay predictions you saved earlier:

python examples/replay.py --save out.mp4

That runs the refiner over the saved frames and writes an annotated video, so you can see what the refinement does without loading a model or spending a call.

How it works

flowchart LR
    P["primary
    detect and track
    every frame"]
    G{"confident
    enough?"}
    H{"voted on this
    track id before?"}
    S["secondary VLM
    the doubtful crops,
    chunk per request"]
    C[("vote cache
    track id to classes")]
    A["write the class"]

    P -- "boxes, ids" --> G
    G -- "yes" --> A
    G -- "no" --> H
    H -- "yes, every later frame" --> A
    H -- "no, once per track" --> S
    S -- "one vote per crop" --> C
    C -- "majority" --> A
    H -. "lookup" .-> C

    classDef vlm stroke:#d97706,stroke-width:3px
    classDef store stroke:#16a34a,stroke-width:2px
    class S vlm
    class C store
    linkStyle 4,5,6 stroke:#d97706,stroke-width:2px

Only the amber box costs real time, and only the bottom path reaches it. A track takes that path on the frame it first looks doubtful and never again, because the answer is filed under its id.

Each frame goes through three steps.

1. The primary detects and tracks. You get boxes, confidences, classes and track ids.
2. Tracks at or below conf are handed to the secondary. Everything above it is left alone.
3. The secondary’s answer is recorded as a vote against the track id. The running majority overwrites the class on this frame and on every later frame, whether or not the secondary runs again.

Step 3 is what makes this affordable. A car the detector is unsure of, in view for 300 frames, costs one VLM call, not 300. A car it is sure of costs none.

There are three modes.

mode="full" runs the secondary on the whole frame, matches its boxes to the tracks by IoU, and takes its box, confidence and class. Use it when the secondary localises: Florence-2, a heavier YOLO, any open-vocabulary detector.

mode="crop" cuts each low confidence track out of the frame and asks the secondary what it is. The box stays as the primary drew it and only the class changes. Use it when the secondary classifies but does not localise, which is every chat VLM.

mode="collage" gathers several crops of the same track, spaced out over time, tiles them into one image and asks about that once. Use it when one frame is not enough to tell, so an attribute rather than a class.

The refiner never adds or drops a box. It only rewrites the class, and in full mode the box and confidence too.

API

vz.Vizor(primary, secondary=None, conf=0.5, mode="full", names=None, **kw)

• conf sends tracks at or below this confidence to the secondary.
• mode is "full", "crop" or "collage".
• names maps class ids to strings. Defaults to whatever the primary reports.
• iou is the minimum overlap to match a secondary box to a track, full mode only.
• votes is how many answers to collect per track before the secondary stops being asked, crop and collage modes. Default 1.
• size and hist cap the vote cache at that many track ids and that many votes each.
• workers runs the secondary on that many background threads, so the frame loop does not wait for it. The track keeps the primary’s label until the answer arrives, then the vote corrects it. Default 0, which blocks.
• samples and every are how many crops make a collage and how many frames apart they are taken, collage mode only. Defaults 4 and 5.
• cell and cols are the collage cell size and column count. cell takes one number for a square or a (width, height) pair.
• chunk, on VLM, is how many crops go in one request. Default 8.

viz.run(src, save=None, show=False) yields refined Tracks for every frame of a file, a camera index, or a stream url. viz.step(img) does one frame. viz.save(src, out) runs the whole thing and writes the annotated video. viz.reset() clears the votes and the primary’s tracker between videos. viz.wait() blocks until the background workers are done, viz.close() stops them, and Vizor is a context manager so a with block closes for you.

Tracks wraps a float32 array of shape (N, 7) holding [x1, y1, x2, y2, conf, cls, id], with id = -1 for untracked boxes:

out.data              # the raw (N, 7) array
out.boxes             # (N, 4) xyxy, a view, so writing to it edits data
out.conf              # (N,) confidences, also a view
out.cls, out.ids      # (N,) ints, copies, so writing to them changes nothing
out[0]                # one Track
out[out.conf > 0.8]   # a new Tracks holding a copy of the matching rows
out.draw()            # draw the boxes on the frame it came from, return the image

Indexing with an int gives you one Track. Indexing with a mask or a slice gives you a new Tracks, and because numpy copies on fancy indexing, writing to it does not touch the original.

Bundled models. Pkl can also be the primary. For the rest the primary is yours to bring.

Writing your own model

Subclass vizor.Model and implement the method your role needs. A primary implements track, a full-mode secondary implements find, a crop-mode secondary implements name. Anything you leave out raises on the first frame with a message saying which role is missing, rather than silently doing nothing.

import numpy as np
from vizor import Model, Tracks

class MyDetector(Model):
    # turns class ids into labels when drawing, and becomes the menu the VLM picks from
    names = {0: "person", 1: "car"}

    def track(self, img):
        # img is BGR, the layout OpenCV hands you. Return one row per object as
        # [x1, y1, x2, y2, conf, cls, id], with id = -1 for anything untracked.
        # An untracked row is refined on this frame and never cached.
        return Tracks(np.zeros((0, 7), np.float32), names=self.names)

Images handed to your model are BGR, the layout OpenCV gives you. Convert inside your wrapper if the model wants RGB. name returns a class id, or None if the model is not sure, and None records no vote.