From 7f3dd13d03f073846f10709dab953e92c584285a Mon Sep 17 00:00:00 2001
From: Heberto Mayorquin <h.mayorquin@gmail.com>
Date: Sat, 17 Sep 2022 03:03:40 +0200
Subject: [PATCH] Add NWB support III - provenance writing (#17)

* added tests for provenance writting

* MyPy compliance
---
 sleap_io/io/nwb.py                            |  79 ++++++++++++++----
 ...ictions_1.2.7_provenance_and_tracking.slp} | Bin 102157 -> 104237 bytes
 tests/fixtures/slp.py                         |   2 +-
 tests/io/test_nwb.py                          |  33 ++++++++
 4 files changed, 95 insertions(+), 19 deletions(-)
 rename tests/data/slp/{tutorial_predictions_version_1.2.7_with_provenance.slp => predictions_1.2.7_provenance_and_tracking.slp} (51%)

diff --git a/sleap_io/io/nwb.py b/sleap_io/io/nwb.py
index cddfd848..30768535 100644
--- a/sleap_io/io/nwb.py
+++ b/sleap_io/io/nwb.py
@@ -109,6 +109,21 @@ def write_labels_to_nwb(
         https://pynwb.readthedocs.io/en/stable/pynwb.file.html#pynwb.file.NWBFile
 
         Defaults to None and default values are used to generate the nwb file.
+
+        pose_estimation_metadata (dict): This argument has a dual purpose:
+
+        1) It can be used to pass time information about the video which is
+        necessary for synchronizing frames in pose estimation tracking to other
+        modalities. Either the video timestamps can be passed to
+        This can be used to pass the timestamps with the key `video_timestamps`
+        or the sampling rate with key`video_sample_rate`.
+
+        e.g. pose_estimation_metadata["video_timestamps"] = np.array(timestamps)
+        or   pose_estimation_metadata["video_sample_rate] = 15  # In Hz
+
+        2) The other use of this dictionary is to ovewrite sleap-io default
+        arguments for the PoseEstimation container.
+        see https://github.com/rly/ndx-pose for a full list or arguments.
     """
 
     nwb_file_kwargs = nwb_file_kwargs or dict()
@@ -129,7 +144,6 @@ def write_labels_to_nwb(
     )
 
     nwbfile = NWBFile(**nwb_file_kwargs)
-
     nwbfile = append_labels_data_to_nwb(labels, nwbfile, pose_estimation_metadata)
 
     with NWBHDF5IO(str(nwbfile_path), "w") as io:
@@ -145,12 +159,33 @@ def append_labels_data_to_nwb(
         labels (Labels): A general labels object
         nwbfile (NWBFile): And in-memory nwbfile where the data is to be appended.
 
+        pose_estimation_metadata (dict): This argument has a dual purpose:
+
+        1) It can be used to pass time information about the video which is
+        necessary for synchronizing frames in pose estimation tracking to other
+        modalities. Either the video timestamps can be passed to
+        This can be used to pass the timestamps with the key `video_timestamps`
+        or the sampling rate with key`video_sample_rate`.
+
+        e.g. pose_estimation_metadata["video_timestamps"] = np.array(timestamps)
+        or   pose_estimation_metadata["video_sample_rate] = 15  # In Hz
+
+        2) The other use of this dictionary is to ovewrite sleap-io default
+        arguments for the PoseEstimation container.
+        see https://github.com/rly/ndx-pose for a full list or arguments.
+
     Returns:
         NWBFile: An in-memory nwbfile with the data from the labels object appended.
     """
 
     pose_estimation_metadata = pose_estimation_metadata or dict()
 
+    # Extract default metadata
+    provenance = labels.provenance
+    default_metadata = dict(scorer=str(provenance))
+    sleap_version = provenance.get("sleap_version", None)
+    default_metadata["source_software_version"] = sleap_version
+
     labels_data_df = _extract_predicted_instances_data(labels)
 
     # For every video create a processing module
@@ -162,20 +197,27 @@ def append_labels_data_to_nwb(
             processing_module_name, nwbfile
         )
 
+        # Propagate video metadata
+        default_metadata["original_videos"] = [f"{video.filename}"]  # type: ignore
+        default_metadata["labeled_videos"] = [f"{video.filename}"]  # type: ignore
+
+        # Overwrite default with the user provided metadata
+        default_metadata.update(pose_estimation_metadata)
+
+        # For every track in that video create a PoseEstimation container
         name_of_tracks_in_video = (
             labels_data_df[video.filename]
             .columns.get_level_values("track_name")
             .unique()
         )
 
-        # For every track in that video create a PoseEstimation container
         for track_index, track_name in enumerate(name_of_tracks_in_video):
             pose_estimation_container = build_pose_estimation_container_for_track(
                 labels_data_df,
                 labels,
                 track_name,
                 video,
-                pose_estimation_metadata,
+                default_metadata,
             )
             nwb_processing_module.add(pose_estimation_container)
 
@@ -260,7 +302,6 @@ def build_pose_estimation_container_for_track(
     )
 
     # Arrange and mix metadata
-
     pose_estimation_container_kwargs = dict(
         name=f"track={track_name}",
         description=f"Estimated positions of {skeleton.name} in video {video_path.name}",
@@ -268,11 +309,7 @@ def build_pose_estimation_container_for_track(
         nodes=skeleton.node_names,
         edges=np.array(skeleton.edge_inds).astype("uint64"),
         source_software="SLEAP",
-        original_videos=[f"{video.filename}"],
-        labeled_videos=[f"{video.filename}"],
         # dimensions=np.array([[video.backend.height, video.backend.width]]),
-        # scorer=str(labels.provenance),
-        # source_software_version=f"{sleap.__version__}
     )
 
     pose_estimation_container_kwargs.update(**pose_estimation_metadata)
@@ -302,28 +339,32 @@ def build_track_pose_estimation_list(
 
     pose_estimation_series_list: List[PoseEstimationSeries] = []
     for node_name in name_of_nodes_in_track:
-        # Add predicted instances only
 
-        data_for_node = track_data_df[
-            node_name,
-        ]
+        # Drop data with missing values
+        data_for_node = track_data_df[node_name].dropna(axis="index", how="any")
 
-        data_for_node_cleaned = data_for_node.dropna(axis="index", how="any")
-        node_trajectory = data_for_node_cleaned[["x", "y"]].to_numpy()
-        confidence = data_for_node_cleaned["score"].to_numpy()
+        node_trajectory = data_for_node[["x", "y"]].to_numpy()
+        confidence = data_for_node["score"].to_numpy()
+
+        reference_frame = (
+            "The coordinates are in (x, y) relative to the top-left of the image. "
+            "Coordinates refer to the midpoint of the pixel. "
+            "That is, t the midpoint of the top-left pixel is at (0, 0), whereas "
+            "the top-left corner of that same pixel is at (-0.5, -0.5)."
+        )
 
         pose_estimation_kwargs = dict(
             name=f"{node_name}",
             description=f"Sequential trajectory of {node_name}.",
             data=node_trajectory,
             unit="pixels",
-            reference_frame="No reference.",
+            reference_frame=reference_frame,
             confidence=confidence,
             confidence_definition="Point-wise confidence scores.",
         )
 
-        # Add timestamps or rate if timestamps are uniform
-        frames = data_for_node_cleaned.index.values
+        # Add timestamps or only rate if the timestamps are uniform
+        frames = data_for_node.index.values
         timestamps_for_data = timestamps[frames]
         sample_periods = np.diff(timestamps_for_data)
         if sample_periods.size == 0:
@@ -334,6 +375,8 @@ def build_track_pose_estimation_list(
             rate = 1 / sample_periods[0] if uniform_samples else None
 
         if rate:
+            # Video sample rates are ints but nwb expect floats
+            rate = float(int(rate))
             pose_estimation_kwargs.update(rate=rate)
         else:
             pose_estimation_kwargs.update(timestamps=timestamps_for_data)
diff --git a/tests/data/slp/tutorial_predictions_version_1.2.7_with_provenance.slp b/tests/data/slp/predictions_1.2.7_provenance_and_tracking.slp
similarity index 51%
rename from tests/data/slp/tutorial_predictions_version_1.2.7_with_provenance.slp
rename to tests/data/slp/predictions_1.2.7_provenance_and_tracking.slp
index 611015e779bceecd364bd2382fdca1bfb23aef07..00c31bd21ed9e0b8df71f3d7a8b626fb3505a5f7 100644
GIT binary patch
delta 7127
zcmZ`-30PD|wx+t-L=ll)fyP$Dq6iIwEZtp3L?X(rh!}R7O`s8w#YRD2bj+wIUbSLS
z;})H0B*Bhij0whhI-_~0(JYQWXEaXC@~L6In3(M)ukO9)kZ!)%eBU{zPW|VvQ&p$x
z-pi&v_Q!hcw*)CdS#QV=g(7@nS8u*KKzUW6^l2wuy?(+Ndq-t@l4Qo%hbw;2boKf>
z<=Du-mm+yfc+=id5ug^W8}_G0+?go%c?&xo5=f_T+aZK>vaS~5uGCgf5LAw@k<PAO
zFX4`(H|gxlm_@#D5blie_6n;o))>Mn48?}}Mq}7)V{>h_v8c2`nBZD4Zb|r5bz^-|
z$<l&wb>8m{jXKXvM&$5#?*O6@-gg*JocVbE&VxR34hpWEt;Zh6$M&gpmHSbOIPvF<
z`FL*>lQp-GizMR*oo#s+v9Ysp<bPES###~TMr>>@2?OKzkJd!+<cjLc4UOtJO}Jq4
zZB}R^gexkQ(C+Uc1p2v=XyG;g>4MSEeN^=C&~Twk6(pSUb9c6IYBZ9U_*+4zOU`6}
zcg}R#Z@RF}U*z=ij}&f8KAmd15Tp`)TE9r4P9@^%e7%J|sxWS?<C#2CP4w49tek36
zSC*>dqR|LPgOBSQFfnWfwj<Hdpb>?JW+*6DSRLSHcAnrzgrfn@LU%xgIYj-OoxQrO
z+SsfXngZ1%Lf93kvTdwiZV>tdCo03I#R&HU17(M6&;$}LL<hxBQdibCG#1sC7}T)M
zm9^#5YKj_5D%5f6@_K`zv{fx^2=XD3!e4lhMi>b4<ML6EXMFgyD2o_lSwVAUZK<)j
zprLYwK^+&NvF@xXYAL8TG#RQJ)NzqQV(@ZeBiyz3XW^7Mn(qW}Bhs!nPkD<d>Y{|u
zP?c1&X`!0-y*hEzZET_^$U2E{zvb_fe7S}Fomf%;8r5Ol|FIG}K+W1`-S2B>)!%N@
za?To1Vs8bo>A7d!f5gGM|KGN9#+cxL9IgAuJ!3s1>>nq=%&VrQj57OaGl}_wI4+F3
zEl%c)x!n3KDT|m7CLd6kcc$`YQVJJ$v$)utDT#v$^V)1~MK0vHki{9q{OKazH05%F
zyfSW$G;r}oImf-alv`O<T+FTI;z>0Z59_$-HgeI`zy+=6;$9<<J>S9k8rwOa*P6Dm
z=FhkD4q+P?GuZb^bNeoCg?4kuwcU}<=0&gbrrl94e(vGEBgc4H%A0)1gtxfwKgYTF
z;v^T_PjIo}Z7yO@adG=JAE^HVkElJ*BV7O9<z#+*om=~^@m}@`7l|M9-mgF1>}3An
z9&fgO#+&}1f9hgR`kptxd(6e=pSjTeB#G(HeSdW%nWS%-1IZ$No=)U|qHpAA(ew5w
zQNA%olzA?q%yJVY87s;s9wPpbJINq@-QJ>Su9xU}<RdcN^AkN?zM_m$kp~Wa4MAeB
zI#9%)4<@fF`p$%kcI8wNQxGmn=P*$!BSqP(5#`}1Q7)V=%6Ty&!%xv-)b3a^$I&?`
zSuT-_yd+|!Th*3dk|n~Mj@~R_7V%c(WcJ0)B5sasZZa9g7AKQXsp#0BlS!29`(ViT
z3;L9QyB5c0&Lbf%BKvRfv)@|4TDemWdt)9_-I<3}6Xv5Yb3Xcn`NTbW;2-z(BcZs7
z<uS74@BX-rZU)dU(4V#*)Q`eI@p!RklZiXCOCj$1-8T}bvt=LH{JHRZ(&G-Tpszv4
zNzi+Y&(hIs3+I?*3FS8)MaW+PdU|u6j%`jM9@3;I+uW|9*mV^kY0KT?$AC@56D+6r
z6{ok2Y1^)2B$ar`i#{%u_$D9S=%gKs_z_@;$}$({c^&j?=z9YCe2JsZ4QwI?N&_jc
zMJF(p3frPpRyI($R%~2FM*bt0*K0jw(}p;KA$4$pyL;PbB^!}OLgY+mq!E9Y4(f$b
z;^xWoYDpuWiVn)Qq`{*2B)6B+7#J!MbXU_}oe%3uBkmf@X>C1SN)eI?(Blnt)8!kD
zXDaTeG>(?eUqlg708sDBmmJ1pO1Sl6+j5Nif3If+={S*1=~#%D(@CgYx(*q9Yn&1X
z9hfQuw<ab7ajh9x#7DSKUXFW1y^}IYsMV=>siTb*X2S9-SIntlJF&&CaIi$MBMaD-
z5udMSv$KYstewV6wvs!g^t4Vducc~uR)v91*fl<W3eFt&a`?<`4w=|DLp1LmOQc!|
zp9}g<`urK8NaG2{aGI9I3s}(toYC2~I`%SeSxwq?rioog2ph2wS1fKJuA*tk_xh0U
z;gHWi8*v5M7`ui0CR=V~TW=n{0f~PO`b_k?*jUz^&F=v5LQ`tDE@RG%AXk>K2+0~3
z;Q%LweD}Fe-cD5x@wN5J)sAE|2dbDA=3sg3<d%%Pl!GiI7o$(R7<`&kV>4}pcZSuV
zBl>PC5^;gm!xotJ@K~$!us$cBc0y=3=>Gj1osy7ZA~)o77nBS6DQ~W+W2L#nc4TTv
zBiqfLa)n>dCGLKEJog9AL!xJf=VH4A=a#qxm@^7`OX*jT#`mCOv~@HUSptW#;N>yU
zQ%EsDDZZNXpU*eH>DiSfc)-3~LVWFfZq29Z?CKKOJ{J5uKZz;j&r$a5^SBYk+~-%r
zFPu!7$A$!hAu6#znD9LA<o`TRd|eNJ<m9{nFeF3-NDm3;Jlr#Vo^+dKBSt()GNwI`
zcq%*SBKBb(=5&f%w#KU;y0NQynEY3Huvl6*?Q*&nX1En}aMY?9xyX&jOKe+K^W?)t
ztSo=nu61l}VXttfT+A1qEM+&ar75%fFZwo&Z~^q)wC;F#tF<9vFeHw|Xu+@(^v;)4
z>2(Ob1N!*vPF*1@EP&~HAG9l@*!K`HF9sWR?85@$VO6Z%{oxc5KLOIG%%CrZJcx!H
z(z52C+AxhG!~vf7V-sBTB?yfGLsXWl6f`G^Vt_jYj?UhsFSQJm#?U+-Q$i6k0iaX6
zXX_0JjRZqfQrTQC#FBR@!qS*oguc~7zT-o_Z-;!7iV<fhhU21r_^XM&2qW(R{iJWS
zPWl)X2c0o213WJ8J}6{2i}4ZVSi(<6_Th6ub!@O0t8i`!UPo(6_<7i9AFizjHWE9G
zmYGZNJUrl*d<BG-;uP~s`5DPpPl;_pf_QGo_5TV-S>5c5(^}cBQmn{}<JU5G0|~LJ
z1I?N$IvGB7B0wK$WezKlUi>bT=qC$3O%x#%pwa!4?M(>PBUGbksiX**56~CRJf~~3
zgn}(^=R*UZ*V&$5;#MMr4zV3BnM)S~OBv{ozE9AvvV>Z(6_oL3Lf6k4Eo(2st7vZ-
zakon^xU2499o&>33Llo?5|1tCJF~$3*Z4Ka`Ycj%EtOnCIe@jrCsxtJkjY!1-z!tc
zynwN@hQ(sBEc9z;XUnlz9+mS+?_QBf*I`@&H>B0}5$-qA%Rqe#bZTm+?oVuW1!iAZ
z!Kc~&gMqS+3Y_LX-kUtO--Eu0oQY4##c-^Feh177_?msSEsOryGElB8w@Oa>zs=rs
z1NO~%3Ta?e19bs1U(g4at<!I`43tUNilkqKEA^YOFX<_y(Pa%Z7|5bQk1(XsPRl@<
z^a&@mHNE(qj=5Cfg&$PK&t%8fCc1@<slqc^$t~OIuiVtL)f^}tv-~OX%V{3)SApLA
zPdmC5$C>+-1i#5`q8QK&fw$TQbpi$^b3;n7YGQU1MaT;Py&jxSw^>4Q7|Yu<^+XHB
zfG!BUU^MEtV<5gBh;6A}S1pyElshZ|@+$Q&m7bK0xmIH>4KC%iWF9c+*_djqB@ORQ
zP8czu+ldq@V2H{xIa5dzJr7A;1Kp^7OaHQEpgi^gk<_#Bimn^`@Qq|R>7$eu`Y#}R
z0Q#2a7>eJML=2QkJ!&}V`jjcU-AI}S#&FWvKdhiX0+~Jhk7ma1f9m&G2Fj#=s)1|U
z=TKG!bpvn!=-jKH>G!g|eBeozD^Y1!Nf8nX(CqcudgfY-^%zpikNvsp?Q}nmnZXTd
z%*id6Drg*pC4o+#R80>cRD6+=p!uam6%-*002-V#iXOCtf^D@~gC(me1{6Wy`MaL_
zS1~XP$H*MKVx$PE1?Z0BuG&ME&_5buRrCsq0VW9i$)t)s%=Q`aZDoH;9mT%&fLVA#
zr)PVO!`@cD9ot9|u^o`x@73sD!!#CR#`(um>L^0?1N6BBv;K%B^p7&$WT>GS@CF2`
z=2hyCVxaiODXo0_FS$mFkTU?yG+oraZVCOPF>EucDF$4Gz{Bh%-7ySYgi%~uR@IlO
z)f59h@rK9;zqQjI%SgFwy_WGa{W!6jvan@vl6|sfx6*qMHVFE&On>d0ES#gQ3V6#U
zBSpyn0d!Qp5Xbf*8#~8i{1V>F&>qLp7Z1yueY*_bVy4#PBetd<-(vRE^Aiw%dDD!O
zNU($(Qu;pq52~nsgzTROT4#51LLbvK5O=Hho68?9VdER{{pRd#BU^wiP4yfDH9&Yh
z=+j>o=}zN_d0>c2x-Q>o;Q!ERZgtYVgAjZ>M+iQ}b@NTM9XL0F{`GF4{)}ayd`NFz
zDx(P53DAYhy!G!Qv;d6ZW3)yzQw%r+fsyv_>ff^rl*aHDu2)fnm;pMo=BS?aHxdu4
zuW-``B@_{70Xa|?6?2xIZG=;)G!1kx-{r8VT9+Hk=l~#Zg1!{kqdkXd6g?&XpJz6+
zPEKR>w(0+oiGBerk3pYQ9@G6dvK4bfnv^nnedCTM$bv;T@qZDXt~;Q=fI$5C5Zjgq
z<$Pit9qlBe#)JO2_!<4**!!GAew+Raw`{w%Z_~50O?W5&qlrKMcC9Vc_x;#JHW0<d
b4z{j^sDA$k?E5YJAFw~J@4MAPB0T>e&%VOP

delta 6598
zcmZvg30zcF7r^HYvWOul7$`0<#3mRsZ0=oCahDJT1cL!)1{{PLgkc?)(X_PElvmx%
zGF;MMivo={<&s!hpY?Oemd`YkzJ6t2TH)7bi}l@i-#u^c_!EA=^Ugcxf6hJk-23jE
znSJ|xHg5OX9H$O=qIdk$YPBkSOK-6=!t0#c>-Pp>OYayl$j8r1i>NxuN2|UNzomDq
z|1y>Q*OWl+#2z0%^}t9;yWsQ2;GS^V7bkA@oh5XMmwm?zUAWgFgm{^6juprGy(DyP
z=?xY8{lbJU*Qy1=RbR0`C``;roFT>xo97e;m-vbin#U>KtyvAavs;rQgsX-QBIyN{
zx5|(FSJqiA(rCegN_9!R6}3}H5qo08#6$N~63rEzDi%fut94p&MQn<=Gde_4=f$Lo
z*JSHvqNc`B;6hA_XpadA&}kDjw$j=PlT9-vUDQR#izj5jLXr)Mr5XMqYL2W%#HNbR
z#nKEnW5<bSW8<VDVxq^1p>a_|QrZ&j8naO|h1izGMP;Qc_sNQ;18pKGX`oC7MPamU
zMQufi8jICvs?$taE?$YZiC@I~JCDbW5u7>W=XnpZ)K^uBJCZc9l0v3ZTWVNowp83(
zQ)|*2h)Hj=>S2DAy_ysywrB&zFzskLc#<|!RtmLKC*GE&y3}S^p_y`<l@6kBG+T_;
zM%gKTq>V(kQPpOPp2q8ItY(Xi)PEabuCCD68A_{6EJva)Tyq;@Ehj0O^i`$?Qx!`f
z8g=#Jmp)^0)?C5wwd5xSY2P?;$^k*0H9=gQHjeHiYnraZktT1<^ho(SmuwpT@AW+S
z4*Z+cyjP{V-stK7#LLsK+3M*Z?d{?3P%S&;2^f4l0M({Tp8l78J^d&C^7P;E^Ys5Q
z#&bsCA4M_e_VmSfIj7Ig7M#~+67fXFW`F0ZB{_mKB5R4@d?$Oi+WBT4aa80I(OpPx
ziwcM)JV?Zg#e`d0M7a5-<n}i`5!=d0f;J1$5~_(fYb6mU>xfoYON6<Bh|qc>0-A|9
z+(<;Pod|sk5ngR1_UrXT6CWX&Cp~H(?)2VHZk^8&k^jQ4Ud|VG6G!GQGUSaHQv;lh
z2Z$ruNkq7dh*JlN*Yd_LwbO8f+zvZTgzqsT4jv`qmE%Md^bv9EZ8Ff6Gep~XnnWa@
zBex-+lcdKl5}WG_Vk^H)ZU^@h$F*zZ_Sx^q?ey<21Uf7JCdi0e*@CM`E$mjih6n=D
zyS;?}3~_xGB;*LL&Ol*_;7S`Vc@yrIyq|?g@@$wS+d?H-7cR-jQIZTDE6J0irPw18
zl3qDZ$`GKDGI+&Emfm|LnHwuy_jQR{Nxwfy;=j>J%;#wmzadq!cTAGxgo%=jnkvZ;
zCrk3Z>5{CUCdrjEBsmgE8Q!{I$nXpBn<?*-Q)E|6rswh?^28X#9=|X=V}TH+p0vPK
zxIhT@!((y<KTK9oCu{p9?bSSDyQ*jrIn3t!5Bc-y)Z>yIVLYX%<-PM$@K&(k4}ivl
zmO`%Siq@cL-AjeAq0*)uLU%4X@4fMkJRvwn8f4oit>_76e~w+H%Uw(td?<?}b6JuA
zPsbQ3$*4Uo=p|-<on58NP~4L%gwXm&OPuO*H3I)T9EZDeg%K2^HJL4S2;Ap5EQsl(
zXHxyxJ5LCc-d`#lnaASseZ)p5I~?G_?#Z4{pM}0-_TSi5x&$*_kXhR6fQ1h1%46N9
zWnCWGh6b@SVHkHdjDTo=IueccGS}hwd=`zD=L@4qLqh=^BG>;UF_Tjf_vEwcX)`+y
z8POz`K7$)(B<{^;&7|cl=rAJiXLFo<K9N-pa8LZ3f7K#j^c>4RTmVZ;M~m`t$1*6r
zcNuK0?-VVn0Q!y!pzjYB0QZ!lol!L3Lh#N2jgHkary4s7`7Qp8qZPji7JLn8e9v;0
zfyQmMnz3;?AD8^~YWzG{Xn$0D>`232I!WBiCjnMrgnPOCpIwIoA7uGxbW_1nG>t{g
z=b|S-^xf|0qx&-OYKW%w4Vz)dhdeBOtu^>3u;Az-=tr@lVMRNqXoHJ^o2h7aMLPf*
z-8a)ZSD_26(yzISC*u!`*}kE*we}X_33^yxy&gt*M$tZ0H2)It&MJX5*O$PWw<_L)
ziuQw|#g;;B323ylnbt;RV=w#F+)F<LI;)YNj`JQ->z5UyO^n{gMa_k%C^>3&85@<(
zHPK+hYs>h3a&b)!J_;86V;MUw7DAdBIZd1aiZ2C?7Jqtw3HmP^{#UMyOo)%A@$|IQ
z_Zkr}Ui6%jA^f8OI(U~+2&12CvhaX`EfUTkM)_DPH3Ha9Z1h;mhJoVcLS05EbO4a0
zfQ*-scoQIHCVCFBPi871K?b%FGnq?1zF~qIJ~lxO116{;pq%BQorvCBiq^4Nb#k*S
zgwZF^<a8N=r`6+!%lTz^_qUb!n1{vD*@SO`g^p0}{SbPSWqy|%p%_L;lFK(#u)64D
zu-m&FeZ=T~E~*4OD%kAk>Ky(y5rO}Ej>9)9*o0{jZ#;G{0{0&rRt81rXc3g%c1(+a
z@#Q|Q8h}l9vp)wJ5HKMeYrF#+H)k#aCX!=Kz^2HJ_}>ax=maxpMn!vG(LPnQz)Ile
zDcWk#=$wvTFF_mFm``(KTVU*obZolJ-~VmLe^kQBlDLxX?HU}k684p8B@DQoxap3r
zg-ny>S^Qf;f~o+XuV`zkSXcvuO_jqA142tV6}}GtS;g<Tf(M#$YBjUasfqg=P&E6}
zkj#B~Xo1YrX=cjm)>I;35Xahqz0b|6o6HE9MI77e$>L!aHkCGTAa|0zhWV&~1($P4
z+C5qP5J^I}C4RvI59$$Oqf6z$T^8v4FBWL#y;e5v8pw6O+`<YgL+F%q$26k7Y~%x6
z1#2N{CXJ%)E<4zc`>n9@!)w?84+A=jqV)L`zp<e+EZrq8?IU<c4b-}pB#<Z7RRfbc
z3Q0FW(hPUEO#AB4FD#jw`=0lxCySF-vC19;#~gQ(*t9wn$bv_6NgnrP@eW9`5gc>f
zNz%*fP&^Ax=aOvlWN~mUOfjhzrnrdM=(=sj+FIzvItAPTAYGmx3mZ@o8=#UKs1pX7
z?=C7~YCQs`kz++*7r5EXgG~sSwH*889oVS61_aC&j(rN)g)}eTSx4VHPv2&~M9k#9
zgR#yAvkkF9TeLQ4%VJ`qCk6XOKqV}8muny2fKIZZ&vPX_3uDPFJ_bp)fg{tM<cGyg
z=n4zI&Lx2_M1ss>T|KMhc}VhrJIU;uE$DX^?8E(*u-%izyXx6fMjr>WsSO1)I)aPp
z!hQ9uDfD%<Cuao${{)W1x(4V|W&^8kCycnnU0u`rE6@}cH<ydw1#GsP9Wbm$z%1q1
z-FIM(r7I9Hr5yVrusQC$?ddHD7z@Yl@nmUV2F_X$Fm{gJi;p+*efep)4S|0>$Kk$4
z)))G;>};w*;C_n3;9IMZ>n?MnsTu*ZlVe|oGC6kN;u-`@H^;sLY@R#spe0oZm?Iqf
z>K)j>3atp3QyjYw*nD^1Z!@b9FrRYleoq#sG{dv?Kr`zNZT05&t>`NjH^9aB;H+l8
z)j6}95%_O$96k=|@E2qV`l!AJV;9gmacB#?eX<m7gQC6N!ouLk0bw}}qvajDUWx7-
zEazLuuF{<gFR`=o=sT$IvK3FY^RN6fA1}u?u+VK#HOGWZEJrPu>kyP%B$s=|&I&pV
zj$(JeFTP)nI#}@IT#_SqV5cv#BVe|2>``F#?!4khl?a$U9Q&3hOOJ!*<VFOHlVkCf
zR(@&iryU5~eH_+@`&;?m**`GgX>I(6qd#XfqSGwmA{Y5K6jMqIr=1VHunO<@h)!Q`
zLswbU&s_8gh%R$SkNwz=BiiA!Lv}ms{JVe}C`z}&$&4B_)L(Xlv#WIX9{#eOO$VPK
t3FOZX{q1lv-U6FzWQXvWpnj<d?{Ns@+`rU)=^($<edcie=@63d{y)Hb-=qKl

diff --git a/tests/fixtures/slp.py b/tests/fixtures/slp.py
index d3088f9c..3774a06f 100644
--- a/tests/fixtures/slp.py
+++ b/tests/fixtures/slp.py
@@ -29,4 +29,4 @@ def slp_predictions():
 @pytest.fixture
 def slp_predictions_with_provenance():
     """The slp file generated with the collab tutorial and sleap version 1.27"""
-    return "tests/data/slp/tutorial_predictions_version_1.2.7_with_provenance.slp"
+    return "tests/data/slp/predictions_1.2.7_provenance_and_tracking.slp"
diff --git a/tests/io/test_nwb.py b/tests/io/test_nwb.py
index ebc13f62..a3f9e119 100644
--- a/tests/io/test_nwb.py
+++ b/tests/io/test_nwb.py
@@ -87,6 +87,39 @@ def test_typical_case_append_with_metadata_propagation(nwbfile, slp_typical):
     assert extracted_dimensions == expected_dimensions
 
 
+def test_provenance_writing(nwbfile, slp_predictions_with_provenance):
+    labels = load_slp(slp_predictions_with_provenance)
+    nwbfile = append_labels_data_to_nwb(labels, nwbfile)
+
+    # Extract processing module
+    video_index = 0
+    video = labels.videos[video_index]
+    video_path = Path(video.filename)
+    processing_module_name = f"SLEAP_VIDEO_{video_index:03}_{video_path.stem}"
+    processing_module = nwbfile.processing[processing_module_name]
+
+    # Test that the provenance information is propagated
+    for pose_estimation_container in processing_module.data_interfaces.values():
+        assert pose_estimation_container.scorer == str(labels.provenance)
+
+
+def test_default_metadata_overwriting(nwbfile, slp_predictions_with_provenance):
+    labels = load_slp(slp_predictions_with_provenance)
+    pose_estimation_metadata = {"scorer": "overwritten_value"}
+    nwbfile = append_labels_data_to_nwb(labels, nwbfile, pose_estimation_metadata)
+
+    # Extract processing module
+    video_index = 0
+    video = labels.videos[video_index]
+    video_path = Path(video.filename)
+    processing_module_name = f"SLEAP_VIDEO_{video_index:03}_{video_path.stem}"
+    processing_module = nwbfile.processing[processing_module_name]
+
+    # Test that the value of scorer was overwritten
+    for pose_estimation_container in processing_module.data_interfaces.values():
+        assert pose_estimation_container.scorer == "overwritten_value"
+
+
 def test_complex_case_append(nwbfile, slp_predictions):
     labels = load_slp(slp_predictions)
     nwbfile = append_labels_data_to_nwb(labels, nwbfile)