fix(smartlab_demo): Detector.inference_async raises TypeError on every call

[Anai-Guo]

Problem

Detector.inference_async in demos/smartlab_demo/python/object_detection/detector.py
cannot complete for any input — it fails twice over:

def inference_multithread(self, img_top, img_side, frame_index):   # L186
    ...
    return [top_bboxes, top_cls_ids, top_labels, top_scores], \
           [side_bboxes, side_cls_ids, side_labels, side_scores], \
           frame_index

async def inference_async(self, img_top, img_side):                # L219
    return await self.inference_multithread(img_top, img_side)     # L220

1. Missing required argument. inference_multithread takes frame_index, but the
   call passes only img_top, img_side →
   TypeError: missing a required argument: 'frame_index'.
2. Awaiting a non-awaitable. inference_multithread is a plain def returning a
   3-tuple, so even with the argument supplied, await raises
   TypeError: object tuple can't be used in 'await' expression.

Why this is the intended shape

The parallel class in the same demo already defines the contract — Segmentor.inference_async
in demos/smartlab_demo/python/segmentor.py:109 is a plain def taking frame_index:

def inference_async(self, frame_top, frame_side, frame_index):

And the demo's own working call site passes all three arguments, confirming the arity:

# demos/smartlab_demo/python/smartlab_demo.py:112
detector.inference_multithread, frame_top, frame_side, frame_counter

Detector.inference_async has no call sites, which is why the breakage has gone unnoticed;
the live path goes through inference_multithread directly.

Verification

Replaying the call against the real signature extracted from the file, with the working
demo call site as a control:

inference_multithread signature: (self, img_top, img_side, frame_index)

UNPATCHED  inference_async -> (self, img_top, img_side)
  TypeError: missing a required argument: 'frame_index'
CONTROL    smartlab_demo.py:112 -> (self, top, side, counter)
  bound OK
PATCHED    inference_async -> (self, img_top, img_side, frame_index)
  bound OK

The second defect reproduced independently:

TypeError: object tuple can't be used in 'await' expression

Fix

Two lines — align Detector.inference_async with its Segmentor sibling:

def inference_async(self, img_top, img_side, frame_index):
    return self.inference_multithread(img_top, img_side, frame_index)

Behaviour of the live inference_multithread path is unchanged.

---

🤖 Generated with Claude Code

[workflow-lab]

Can one of the admins verify this patch?