Datasets:

Dhi-Technologies
/

multicam-reasoning-memory-benchmark

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
t: double
camera_id: string
track_id: string
class: string
position_m: list<item: double>
  child 0, item: double
identity_ref: string
events: int64
linking: struct<precision: double, recall: double, linked_pairs: int64, true_pairs: int64>
  child 0, precision: double
  child 1, recall: double
  child 2, linked_pairs: int64
  child 3, true_pairs: int64
sample_where_is: string
memory: struct<entities: int64, episodes: int64, links_active: int64, relations: int64, db_bytes: int64>
  child 0, entities: int64
  child 1, episodes: int64
  child 2, links_active: int64
  child 3, relations: int64
  child 4, db_bytes: int64
sample_journeys: string
to
{'events': Value('int64'), 'linking': {'precision': Value('float64'), 'recall': Value('float64'), 'linked_pairs': Value('int64'), 'true_pairs': Value('int64')}, 'memory': {'entities': Value('int64'), 'episodes': Value('int64'), 'links_active': Value('int64'), 'relations': Value('int64'), 'db_bytes': Value('int64')}, 'sample_where_is': Value('string'), 'sample_journeys': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 478, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2818, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              t: double
              camera_id: string
              track_id: string
              class: string
              position_m: list<item: double>
                child 0, item: double
              identity_ref: string
              events: int64
              linking: struct<precision: double, recall: double, linked_pairs: int64, true_pairs: int64>
                child 0, precision: double
                child 1, recall: double
                child 2, linked_pairs: int64
                child 3, true_pairs: int64
              sample_where_is: string
              memory: struct<entities: int64, episodes: int64, links_active: int64, relations: int64, db_bytes: int64>
                child 0, entities: int64
                child 1, episodes: int64
                child 2, links_active: int64
                child 3, relations: int64
                child 4, db_bytes: int64
              sample_journeys: string
              to
              {'events': Value('int64'), 'linking': {'precision': Value('float64'), 'recall': Value('float64'), 'linked_pairs': Value('int64'), 'true_pairs': Value('int64')}, 'memory': {'entities': Value('int64'), 'episodes': Value('int64'), 'links_active': Value('int64'), 'relations': Value('int64'), 'db_bytes': Value('int64')}, 'sample_where_is': Value('string'), 'sample_journeys': Value('string')}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Multicam Reasoning Memory Benchmark (E1)

Product: multicam-reasoning-memory ("fleetmind"), "your cameras think together and remember": cross-camera identity linking with transit-time priors, weeks-scale bounded memory, and provenance-carrying queries.

Synthetic data, real-data validation in progress

This is a procedurally generated 4-camera site: simulated people walk a camera chain with per-camera dwell plus transit jitter, emitting per-camera track events exactly like a real per-camera tracker would (new track ID each time a person reappears on a different camera). Ground truth (which track IDs belong to the same physical person) ships alongside it so linking precision and recall are measured, not assumed. No real multi-camera footage or real people are involved, and real-site validation has not been performed yet.

What's in this dataset

events.jsonl: 1,832 track events from a synthetic 4-camera site, 20 simulated people (seed=1), 40% carrying an external identity_ref (stand-in for a plate/face-free cross-link from a sister product). Schema (one JSON object per line):
```
{"t": 28.84, "camera_id": "cam0", "track_id": "t27", "class": "person",
 "position_m": [9.57, 16.40], "identity_ref": "ref:person8"}
```
ground_truth.json: person_tracks (which camera:track_id keys belong to the same physical person), the true mean inter-camera transit times, and generation metadata.
demo_seed1_report.json: the repo's own committed evidence file, a linking plus memory-footprint report for this exact seed.

Measured result (from this repo, `fleetmind.cli demo --seed 1`)

Linking: precision 1.0, recall 0.377 (69 true same-person track pairs; 26 linked, 0 wrong)
Memory footprint for the entire 4-camera site: 61,440 bytes (46 entities, 61 episodes)
Sample journey query (cam0 to cam1): recovered real transits from 37.85 s to 53.65 s, with episode-id provenance attached to every answer

Recall is intentionally moderate: the linker only links pairs it is confident about (precision 1.0, it makes zero wrong links in this run); people without an identity_ref and with transit times far from the learned prior are left unlinked rather than guessed.

Reproduce with: PYTHONPATH=src .venv/bin/python -m fleetmind.cli demo --seed 1

Method card, no trained weights

No trained model or weights ship with this product. Cross-camera linking is a transit-time prior plus a scored decision with a uniqueness guard (a refusal gate: if a second candidate is even plausible, the link is refused). On this synthetic site, margin alone gave 0.918 precision; the uniqueness guard took it to 1.0 by trading recall, a deliberate, measured choice, since a false merge poisons weeks of memory while a missed link only costs a duplicate row. Read the deep-dive post.