Datasets:

RyanIRL
/

ca1-cross-day-decoding

frame int64	t_s float64	head_x_cm float64	head_y_cm float64	speed_cm_s float64	reliable int64	scored int64
0	0	5.671	34.625	0.595	0	0
1	0.069	5.671	34.624	0.668	0	0
2	0.1379	5.669	34.619	0.81	0	0
3	0.2069	5.667	34.601	1.01	0	0
4	0.2759	5.661	34.555	1.252	0	0
5	0.3448	5.658	34.462	1.519	0	0
6	0.4138	5.667	34.311	1.79	1	0
7	0.4828	5.708	34.115	2.047	1	1
8	0.5517	5.799	33.908	2.274	1	1
9	0.6207	5.937	33.733	2.458	1	1
10	0.6897	6.102	33.625	2.593	1	1
11	0.7586	6.267	33.609	2.675	1	1
12	0.8276	6.416	33.702	2.707	1	1
13	0.8966	6.549	33.896	2.694	0	0
14	0.9655	6.667	34.149	2.645	1	1
15	1.0345	6.769	34.381	2.573	1	1
16	1.1034	6.843	34.516	2.493	1	1
17	1.1724	6.875	34.523	2.419	1	1
18	1.2414	6.863	34.425	2.365	1	1
19	1.3103	6.817	34.277	2.336	1	1
20	1.3793	6.752	34.129	2.334	1	1
21	1.4483	6.672	34.006	2.352	1	1
22	1.5172	6.569	33.9	2.377	1	1
23	1.5862	6.441	33.784	2.394	1	1
24	1.6552	6.3	33.629	2.391	1	1
25	1.7241	6.167	33.433	2.356	1	1
26	1.7931	6.068	33.223	2.286	1	1
27	1.8621	6.016	33.044	2.185	1	1
28	1.931	6.009	32.92	2.061	0	0
29	2	6.039	32.846	1.928	1	0
30	2.069	6.092	32.786	1.799	1	0
31	2.1379	6.156	32.707	1.687	0	0
32	2.2069	6.217	32.609	1.605	1	0
33	2.2759	6.267	32.519	1.558	1	0
34	2.3448	6.312	32.46	1.552	1	0
35	2.4138	6.363	32.432	1.589	0	0
36	2.4828	6.432	32.415	1.668	1	0
37	2.5517	6.526	32.387	1.785	0	0
38	2.6207	6.644	32.343	1.934	1	0
39	2.6897	6.781	32.292	2.107	1	1
40	2.7586	6.936	32.25	2.295	1	1
41	2.8276	7.108	32.223	2.488	1	1
42	2.8966	7.3	32.21	2.675	1	1
43	2.9655	7.51	32.196	2.848	1	1
44	3.0345	7.731	32.165	2.997	1	1
45	3.1034	7.957	32.102	3.115	0	0
46	3.1724	8.179	31.996	3.197	1	1
47	3.2414	8.396	31.851	3.238	1	1
48	3.3103	8.598	31.68	3.24	1	1
49	3.3793	8.774	31.495	3.205	1	1
50	3.4483	8.907	31.302	3.141	1	1
51	3.5172	8.989	31.107	3.057	1	1
52	3.5862	9.022	30.915	2.963	1	1
53	3.6552	9.021	30.734	2.868	1	1
54	3.7241	8.997	30.571	2.777	1	1
55	3.7931	8.944	30.428	2.693	1	1
56	3.8621	8.848	30.312	2.61	1	1
57	3.931	8.696	30.22	2.519	1	1
58	4	8.494	30.147	2.413	1	1
59	4.069	8.276	30.078	2.282	1	1
60	4.1379	8.085	30.002	2.122	1	1
61	4.2069	7.949	29.919	1.936	1	0
62	4.2759	7.871	29.835	1.731	1	0
63	4.3448	7.83	29.766	1.519	1	0
64	4.4138	7.801	29.719	1.313	1	0
65	4.4828	7.768	29.696	1.126	1	0
66	4.5517	7.732	29.69	0.968	1	0
67	4.6207	7.697	29.692	0.845	1	0
68	4.6897	7.665	29.692	0.758	1	0
69	4.7586	7.631	29.688	0.706	1	0
70	4.8276	7.591	29.682	0.685	1	0
71	4.8966	7.55	29.679	0.69	1	0
72	4.9655	7.515	29.686	0.716	1	0
73	5.0345	7.493	29.713	0.758	1	0
74	5.1034	7.483	29.763	0.81	1	0
75	5.1724	7.483	29.831	0.868	1	0
76	5.2414	7.494	29.905	0.927	1	0
77	5.3103	7.524	29.971	0.982	1	0
78	5.3793	7.576	30.02	1.03	1	0
79	5.4483	7.65	30.051	1.068	1	0
80	5.5172	7.737	30.065	1.095	1	0
81	5.5862	7.826	30.071	1.11	1	0
82	5.6552	7.913	30.075	1.113	1	0
83	5.7241	7.994	30.083	1.105	1	0
84	5.7931	8.071	30.093	1.089	1	0
85	5.8621	8.144	30.101	1.067	1	0
86	5.931	8.213	30.106	1.043	0	0
87	6	8.279	30.109	1.017	0	0
88	6.069	8.344	30.112	0.993	0	0
89	6.1379	8.408	30.115	0.971	0	0
90	6.2069	8.471	30.119	0.953	0	0
91	6.2759	8.534	30.123	0.938	0	0
92	6.3448	8.597	30.127	0.927	0	0
93	6.4138	8.66	30.131	0.923	0	0
94	6.4828	8.724	30.135	0.927	0	0
95	6.5517	8.787	30.14	0.943	0	0
96	6.6207	8.849	30.143	0.976	0	0
97	6.6897	8.909	30.145	1.03	0	0
98	6.7586	8.961	30.142	1.107	0	0
99	6.8276	8.994	30.127	1.206	0	0

End of preview. Expand in Data Studio

ca1-crossday-decoding

Data for the terminal-bench-science task ca1-crossday-decoding: decode a mouse's position online on a new recording day from raw two-photon calcium imaging of hippocampal CA1, given a full labeled session from days earlier and the first minutes of the new day.

Source

Zong, W., Obenhaus, H. A., Skytoen, E. R., et al. (2022). Large-scale two-photon calcium imaging in freely moving mice. Cell 185(6), 1240-1256. Deposited in the NIRD Research Data Archive, DOI 10.11582/2022.00008, licence CC BY 4.0. Mouse 97289, imaged in CA1 with the MINI2P miniature two-photon microscope while foraging in a square open field, from the deposit's ca1-97289 tarball: sessions 2021-03-16 (day A) and 2021-03-22 (day B), each 26,100 raw ScanImage pages (int16, 256 x 256, the focus alternating between two planes on every page, 14.5 frames per second) with their DeepLabCut tracking.

What was done

20210316_to_20210322/ holds one day pair.

Frames are written as acquired, interleaved and unregistered, to plain TIFFs without the ScanImage headers.
Positions are the midpoint of the two ears in arena centimetres: origin at the bottom-left corner of the box, computed per session from the four tracked box corners, so both days share one frame although the camera moved. Frames with low tracker confidence or impossible geometry are marked reliable = 0 and interpolated, positions are lightly smoothed, and scored marks reliable frames moving faster than 2 cm/s.
agent/ is what the agent trains on: all of day A, and day B's first 5 minutes (day_b_calibration/). judge/ is the rest of day B (day_b_stream/), which the verifier streams to the submitted decoder and grades on, and split.json, where its stretch before minute 15 of day B ends.

Downloads last month: 327