Oasis

Oasis is Cactus Compute's Mirage with a new CPU engine and a retrained text reader. Mirage turns a photo into a JSON record of the scene, the objects in it with boxes and masks, the text it can read, how the objects relate and a short reasoning trace; given a form, it fills in choices, counts, text and boxes. Oasis keeps Mirage's 22 MB of mostly 2-bit weights, runs them several times faster, reads more of the text, and always returns valid JSON.

Oasis and Mirage

Both models described the same photos with the same options, Mirage on its own runtime. results/comparison.json holds every figure with the hashes of both runtimes and weight files.

	Mirage	Oasis
Describe a photo	891 ms	296 ms
Read the text in a photo	917 ms	101 ms
Fill a four-field form	729 ms	164 ms
Records that are not valid JSON, of 300 COCO photos	58	0
Empty records at conf 0.5, of 300 COCO photos	2	0
ICDAR2013 words read	80.9%	81.6%
IIIT5K words read	77.6%	78.7%
Accented letters read	0.7%	30.4%
COCO objects found, F1	0.273	0.294
COCO mask IoU	0.670	0.680
COCO outline IoU	0.673	0.676

Timings are medians over 24 photos. Words are read end to end: each word image is placed in a scene, found by the engine's own text detector and read. Accented letters are counted on 500 synthetic lines in the seven description languages, set in fonts the reader never trained on. Objects are scored against COCO val2017 boxes at IoU 0.5, and masks and filled outlines against COCO's instance outlines.

What changed

The engine is new. It multiplies Mirage's 2-bit weights in place with integer kernels, keeps its worker threads awake while the decoder writes, and skips the decoder when only text is requested. Every record it writes parses. On about one photo in five, Mirage's runtime adds a phantom box in the top-right corner, as an object or as a line of text, with a nan score that breaks the JSON; the same phantom box becomes the answer to box questions in forms. When the decoder's reasoning trace falls into a loop, Mirage's runtime spends the whole token budget on it and returns an empty record; Oasis closes the trace and the description is written. Outlines follow the probability field between mask cells, keywords steer the description toward the words an application cares about, and a language the model was not trained to describe in is refused with an error.

The text reader is retrained. Mirage's reader never learned 67 of its 306 characters, and on accented text cut as tightly as its detector cuts it, it writes almost no accents; Oasis's reader learned from TextOCR word crops and from synthetic lines in the seven description languages, cut as tightly as the detector cuts text, through the same 2-bit quantiser the engine runs, with Mirage's reader as its teacher on real words. The rest of the model is Mirage's, unchanged.

Use

git clone https://huggingface.co/phanerozoic/oasis && cd oasis
cmake -S engine -B engine/build -DCMAKE_BUILD_TYPE=Release && cmake --build engine/build -j
engine/build/oasis_cli oasis.oasis photo.jpg '{"language": "de"}'

import sys
sys.path.insert(0, "engine/python")
from oasis_engine import Oasis

model = Oasis("oasis.oasis")
record = model.describe("photo.jpg", sections=["objects", "text"], mask_format="polygons")
form = model.extract("photo.jpg", {
    "sign": {"enum": ["stop", "yield", "speed limit"]},
    "cars": {"type": "integer", "minimum": 0, "maximum": 10},
    "where": {"x-source": "bbox", "x-of": "sign"},
})

Options and form fields follow Mirage's: language (en, de, fr, es, it, nl, pl), keywords, conf, text_conf, max_objects, max_text_lines, max_tokens, sections, fields, masks, mask_format (largest, polygons, merged or bitmap), mask_threshold, nms and reasoning. Boxes and outlines use a 0–1000 scale on each image axis. The engine builds with GCC, Clang and MSVC.

Files

Path	Contents
oasis.oasis	Engine weights
oasis.safetensors	Float32 weights for further training
engine/	C++ engine, C API, command-line tool and Python binding
reference/	PyTorch forward pass, quantiser, reader training, export and benchmarks
results/comparison.json	The measurements above

The reference tools read Mirage's mirage.safetensors and mirage.cact from reference/weights/mirage/, beside cq_plan.json. Oasis is released under Apache-2.0, as Mirage is; NOTICE records what comes from Mirage, TextOCR and Pillow.